{"about":"Istari: AI coding models ranked by cost per finished task. Written for AI agents; people see the same data at https://istari-index.pages.dev/. Start at https://istari-index.pages.dev/llms.txt.","generatedAt":"2026-10-11T18:05:53.267Z","refreshedEveryHours":3,"schema":{"picks.overall / picks.byTask[key]":"The pick (cheapest measured model you can get in Cursor, on OpenRouter or in a paid plan), the runner-up, and the pick within each tool (see tools). taskScoreEstimated: true when its score for that kind of work is an estimate, allowed only where fewer than 60% of the models are measured on it.","picks.byPlan[plan].overall / .byTask[key]":"The pick within each paid plan (see plans): model, level and costUsd (vendor and basis are in models[]). On a \"limits\" plan tokens come with the plan, so cost is only the time fixing failed attempts, and the pick is the leanest level in the plan within frontier.errorBand of the cheapest; on a \"pool\" plan tokens are billed at list price, as per token.","frontier.overall":"The efficient line of success vs tokens for all coding work: the model and effort pairs no other pair beats on both fewer tokens per finished task and more success, fewest tokens first. tokensUsd is at list price, retries included.","frontier.errorBand / fewestFailures / fewestTokens":"errorBand: the back-tested error of a measured coding score, in points; a smaller gap is not worth more tokens. fewestFailures: the leanest pair within errorBand of the most accurate. fewestTokens: the leanest pair at or above qualityBarSuccess.","plans":"Paid plans that come with coding models: tool, price, billing (\"limits\" or \"pool\"), limits in the vendor's words, model-specific notes and the date checked against the vendor's docs.","models[].cursor / models[].openrouter / models[].plans":"Its name in Cursor, its OpenRouter id and the paid plans that include it. A model in none of these is never the pick.","models[].basis":"measured | rebuilt (◇: measured, computed from its own current Artificial Analysis results since AA stopped publishing its Coding Index) | provisional (new-model estimate) | hand-estimated.","models[].overall":"All coding work: rank, cost per finished task in USD (tokensUsd + fixTimeUsd), coding score 0-100, the effort level it is ranked at, and USD per 1M tokens at that level.","models[].byTask[key]":"The same for one kind of work: rank, cost, the task score used and the level; status \"partial\" or \"estimated\" when that score is not fully measured (absent when measured).","models[].levels":"Every tested effort level: its coding score, price, thinking multiplier (1 = \"high\") and cost per task; status says why a level is not ranked (absent when ranked).","models[].note":"How an estimated or rebuilt score was obtained, with its typical error.","formula":"How a cost per finished task is computed from a score, a price and a level, with the constants."},"formula":{"successChance":"p = score / 100","tokensPerTaskMillions":"tokensPerTaskMillions × (1 − reasoningShare + reasoningShare × thinkingMultiplier)","costPerFinishedTask":"(pricePerMTok × tokensPerTaskMillions_at_level + fixCostUsd × (1 − p)) / p","byTask":"For one kind of work, every level's score shifts by (task score − coding score) of the model, and the cheapest ranked level is used.","constants":{"tokensPerTaskMillions":0.27,"developerHourlyUsd":93,"minutesPerFailedAttempt":20,"reasoningShare":0.4,"fixCostUsd":31}},"qualityBar":{"ratio":0.85,"bestScore":81.8,"bestModel":"Claude Opus 5.5","minScore":69.53},"tools":[{"key":"cursor","label":"Cursor","includes":"models you can pick in Cursor"},{"key":"claude-code","label":"Claude Code","includes":"models in Claude Max"},{"key":"codex","label":"Codex","includes":"models in ChatGPT Pro"},{"key":"antigravity","label":"Antigravity","includes":"models in Google AI Ultra; it replaced Gemini CLI for personal accounts"},{"key":"copilot","label":"GitHub Copilot","includes":"models in Copilot Pro+"}],"plans":[{"key":"claude-pro","family":"Claude","label":"Claude Pro","tier":"Pro","tool":"Claude Code","price":"$20 a month","billing":"limits","limits":"A 5-hour session limit and a weekly limit across all models (Anthropic publishes no token numbers). Past them you wait, or pay usage credits at API prices.","checked":"2026-10-10","notes":{}},{"key":"claude-max","family":"Claude","label":"Claude Max","tier":"Max","tool":"Claude Code","price":"$100 or $200 a month","billing":"limits","limits":"5× or 20× Pro's 5-hour limit, plus a weekly limit across all models. Past them you wait, or pay usage credits at API prices.","checked":"2026-10-10","notes":{"Claude Fable 5.1":"Fable can use up to half of the weekly limit; past that it runs on paid usage credits.","Claude Fable 5":"Fable can use up to half of the weekly limit; past that it runs on paid usage credits."}},{"key":"claude-team","family":"Claude","label":"Claude Team","tier":"Team","tool":"Claude Code","price":"$20 or $25 a user a month","billing":"limits","limits":"More usage than Pro, with a 5-hour and a weekly limit per seat. Past them you wait, or pay usage credits at API prices.","checked":"2026-10-10","notes":{}},{"key":"claude-team-premium","family":"Claude","label":"Claude Team Premium","tier":"Team Premium","tool":"Claude Code","price":"$100 or $125 a user a month","billing":"limits","limits":"5× a standard Team seat, plus a weekly limit. Past them you wait, or pay usage credits at API prices.","checked":"2026-10-10","notes":{"Claude Fable 5.1":"Fable can use up to half of the weekly limit; past that it runs on paid usage credits.","Claude Fable 5":"Fable can use up to half of the weekly limit; past that it runs on paid usage credits."}},{"key":"claude-enterprise","family":"Claude","label":"Claude Enterprise or API key","tier":"Enterprise · API","tool":"Claude Code","price":"Per token, at API prices (Enterprise adds $20 a user a month)","billing":"pool","limits":"Every token at API prices, within the spend limits your admin sets (Enterprise) or your API budget.","checked":"2026-10-10","notes":{}},{"key":"chatgpt-plus","family":"ChatGPT","label":"ChatGPT Plus or Business","tier":"Plus · Business","tool":"Codex","price":"$20 a month ($20–25 a user on Business)","billing":"limits","limits":"A message limit per 5 hours that depends on the model, and a weekly limit. Past them you wait, or buy credits.","checked":"2026-10-10","notes":{"GPT-6 Astra":"On Plus only about 5–45 messages per 5 hours.","GPT-5.5":"Leaves Codex on 2026-10-14."}},{"key":"chatgpt-pro","family":"ChatGPT","label":"ChatGPT Pro","tier":"Pro","tool":"Codex","price":"$100, $200 or $500 a month","billing":"limits","limits":"No 5-hour limit, only a weekly one. Past it you wait, or buy credits. Ultrafast mode (Pro $500 only) uses the limit 8× faster.","checked":"2026-10-10","notes":{"GPT-5.5":"Leaves Codex on 2026-10-14."}},{"key":"chatgpt-enterprise","family":"ChatGPT","label":"ChatGPT Enterprise, Edu or API key","tier":"Enterprise · API","tool":"Codex","price":"Per token, as credits or at API prices (Enterprise seats by quote)","billing":"pool","limits":"Every token paid for, as workspace credits at OpenAI's rate card (Enterprise, Edu) or at API prices (API key), within your admin's or your own budget.","checked":"2026-10-10","notes":{"GPT-6.1 Sol":"On Enterprise and Edu it is off until an admin turns it on.","GPT-5.5":"Leaves Codex on 2026-10-14."}},{"key":"google-ai-pro","family":"Google AI","label":"Google AI Pro","tier":"Pro","tool":"Antigravity","price":"$19.99 a month","billing":"limits","limits":"A quota that refreshes every 5 hours, plus a weekly limit. Past them you wait, or spend AI credits.","checked":"2026-10-04","notes":{"Claude Opus 5.5":"Not on trial subscriptions.","Claude Sonnet 5.5":"Not on trial subscriptions."}},{"key":"google-ai-ultra","family":"Google AI","label":"Google AI Ultra","tier":"Ultra","tool":"Antigravity","price":"$99.99 or $199.99 a month","billing":"limits","limits":"The highest quota, refreshed every 5 hours, plus a weekly limit. Past them you wait, or spend AI credits.","checked":"2026-10-04","notes":{}},{"key":"gemini-enterprise","family":"Google AI","label":"Gemini Enterprise or Google Cloud","tier":"Enterprise","tool":"Antigravity","price":"Per token, at Google Cloud prices (license by quote)","billing":"pool","limits":"Usage billed at Gemini Enterprise consumption prices, beyond any quota the license includes, within your admin's budget.","checked":"2026-10-05","notes":{}},{"key":"cursor","family":"Cursor","label":"Cursor Pro, Pro+ or Ultra","tier":"Pro · Pro+ · Ultra","tool":"Cursor","price":"$20, $60 or $200 a month","billing":"pool","limits":"A monthly allowance billed at each model's API price (Cursor's own models get a larger one); past it, pay as you go if you turn it on.","checked":"2026-10-06","notes":{}},{"key":"cursor-teams","family":"Cursor","label":"Cursor Teams","tier":"Teams","tool":"Cursor","price":"$40 or $120 a user a month","billing":"pool","limits":"Standard or Premium seats (Premium has 5× Standard's agent limits). Models at API prices; third-party ones add Cursor's $0.25 per million tokens.","checked":"2026-10-06","notes":{}},{"key":"cursor-enterprise","family":"Cursor","label":"Cursor Enterprise","tier":"Enterprise","tool":"Cursor","price":"Per token, from a pool the company shares (by quote)","billing":"pool","limits":"Usage pooled across the company and invoiced, at API prices; third-party models add Cursor's $0.25 per million tokens.","checked":"2026-10-06","notes":{}},{"key":"copilot-pro","family":"GitHub Copilot","label":"GitHub Copilot Pro","tier":"Pro","tool":"Copilot","price":"$10 a month","billing":"pool","limits":"$15 of AI credits a month, billed at each model's API price; past them, set a budget or wait for the 1st of the month.","checked":"2026-10-04","notes":{}},{"key":"copilot-pro-plus","family":"GitHub Copilot","label":"GitHub Copilot Pro+ or Max","tier":"Pro+ · Max","tool":"Copilot","price":"$39 or $100 a month","billing":"pool","limits":"$70 (Pro+) or $200 (Max) of AI credits a month, billed at each model's API price; past them, set a budget or wait for the 1st of the month.","checked":"2026-10-04","notes":{}},{"key":"copilot-business","family":"GitHub Copilot","label":"GitHub Copilot Business","tier":"Business","tool":"Copilot","price":"$19 a user a month","billing":"pool","limits":"$19 of AI credits a user a month, billed at each model's API price; past them, your admin can add budget.","checked":"2026-10-05","notes":{}},{"key":"copilot-enterprise","family":"GitHub Copilot","label":"GitHub Copilot Enterprise","tier":"Enterprise","tool":"Copilot","price":"$39 a user a month","billing":"pool","limits":"$39 of AI credits a user a month, billed at each model's API price; past them, your admin can add budget.","checked":"2026-10-05","notes":{}}],"frontier":{"overall":[{"model":"MiMo-V2.6-Flash","level":"standard","success":72.8,"tokensUsd":0.08,"costUsd":11.66},{"model":"Claude Haiku 5.5","level":"xhigh","success":73.9,"tokensUsd":0.13,"costUsd":11.07},{"model":"Ling 3.1 Flash","level":"standard","success":75.2,"tokensUsd":0.22,"costUsd":10.44},{"model":"MiMo-V2.6-Pro","level":"standard","success":77.5,"tokensUsd":0.23,"costUsd":9.23},{"model":"Claude Sonnet 5.5","level":"xhigh","success":78.5,"tokensUsd":2.44,"costUsd":10.93},{"model":"Claude Opus 5.5","level":"medium","success":78.7,"tokensUsd":3.38,"costUsd":11.77},{"model":"Claude Opus 5.5","level":"high","success":79.4,"tokensUsd":3.77,"costUsd":11.81},{"model":"Claude Opus 5.5","level":"xhigh","success":81.2,"tokensUsd":4.6,"costUsd":11.77},{"model":"Claude Opus 5.5","level":"max","success":81.8,"tokensUsd":9.52,"costUsd":16.42}],"errorBand":2.6,"qualityBarSuccess":69.53,"fewestFailures":{"model":"Claude Opus 5.5","level":"high"},"fewestTokens":{"model":"MiMo-V2.6-Flash","level":"standard"}},"taskTypes":[{"key":"greenfield","label":"Build a new service or API","description":"Brand-new code from a spec: an API, a service, a module or a whole app.","pickWhen":"building a new API, service, module or app from a spec","benchmarks":["deepswe","frontierSwe","cursorBench","vibeCode"]},{"key":"legacy","label":"Change existing / legacy code","description":"Adapt, refactor or migrate a large existing codebase.","pickWhen":"refactoring, migrating a framework or version, or changing behavior in a large existing codebase","benchmarks":["codeMigration","frontierCode","sweAtlasRefactor"]},{"key":"explore","label":"Explore & explain a codebase","description":"Trace flows across files and repos, onboard, answer \"how does this work\", review.","pickWhen":"understanding how code works, onboarding, code review, and writing a plan for a change","benchmarks":["sweAtlasQna"]},{"key":"debug","label":"Debug, fix & test","description":"Reproduce, locate and fix bugs; write and repair tests.","pickWhen":"a bug report, a crash or a failing test; writing or repairing tests","benchmarks":["deepswe","terminalBench4","sweAtlasTests"]},{"key":"prod","label":"Production investigation","description":"Alerts, events and traces across services to find the root cause.","pickWhen":"an alert or incident: finding the root cause in logs, events and traces across services","benchmarks":["itbench"]},{"key":"infra","label":"Infra, DevOps & scripting","description":"CI/CD, containers, shell, builds and environment setup.","pickWhen":"CI/CD, containers, shell scripts, builds and environment setup","benchmarks":["terminalBench","terminalBench4"]},{"key":"algorithms","label":"Algorithms, data & scientific code","description":"Numerical code, data processing, algorithms.","pickWhen":"numerical or scientific code, data processing, algorithms","benchmarks":["sciCode","ioi","aleBench","tbScience"]},{"key":"ui","label":"UI & web pages","description":"Front-end pages, components and web apps.","pickWhen":"front-end pages, components, styling and web app UIs","benchmarks":["webdev","vibeCode"]},{"key":"security","label":"Security & vulnerabilities","description":"Find and fix vulnerabilities without breaking the code.","pickWhen":"finding or fixing a vulnerability, a security review, hardening","benchmarks":["cweBench","valsCyber","cyberGym","deepsecBench","valsReverse"]}],"picks":{"overall":{"pick":{"model":"MiMo-V2.6-Pro","vendor":"Xiaomi","level":"standard","costUsd":9.23,"basis":"rebuilt"},"runnerUp":{"model":"Claude Haiku 5.5","vendor":"Anthropic","level":"max","costUsd":10.35,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Haiku 5.5","vendor":"Anthropic","level":"max","costUsd":10.35,"basis":"rebuilt"},"claude-code":{"model":"Claude Haiku 5.5","vendor":"Anthropic","level":"max","costUsd":10.35,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":11.17,"basis":"rebuilt"},"antigravity":{"model":"Gemini 3.8 Flash","vendor":"Google","level":"high","costUsd":10.43,"basis":"measured"},"copilot":{"model":"Claude Haiku 5.5","vendor":"Anthropic","level":"max","costUsd":10.35,"basis":"rebuilt"}}},"byTask":{"greenfield":{"pick":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":11.85,"basis":"rebuilt"},"runnerUp":{"model":"MiMo-V2.6-Pro","vendor":"Xiaomi","level":"standard","costUsd":12.75,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":11.85,"basis":"rebuilt"},"claude-code":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":11.85,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":13.05,"basis":"rebuilt"},"antigravity":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":11.85,"basis":"rebuilt"},"copilot":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":11.85,"basis":"rebuilt"}}},"legacy":{"pick":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.74,"basis":"rebuilt"},"runnerUp":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.91,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.74,"basis":"rebuilt"},"claude-code":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.74,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":22.2,"basis":"rebuilt"},"antigravity":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.74,"basis":"rebuilt"},"copilot":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.74,"basis":"rebuilt"}}},"explore":{"pick":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":21.17,"basis":"rebuilt"},"runnerUp":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":22.32,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":21.17,"basis":"rebuilt"},"claude-code":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":21.17,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":25.33,"basis":"rebuilt"},"antigravity":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":21.17,"basis":"rebuilt"},"copilot":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":21.17,"basis":"rebuilt"}}},"debug":{"pick":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":10.87,"basis":"rebuilt"},"runnerUp":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":11.57,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":10.87,"basis":"rebuilt"},"claude-code":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":10.87,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":11.57,"basis":"rebuilt"},"antigravity":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":10.87,"basis":"rebuilt"},"copilot":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":10.87,"basis":"rebuilt"}}},"prod":{"pick":{"model":"GPT-5.6 Sol","vendor":"OpenAI","level":"high","costUsd":28.95,"basis":"measured"},"runnerUp":{"model":"Gemini 3.8 Flash","vendor":"Google","level":"high","costUsd":29.16,"basis":"measured"},"byTool":{"cursor":{"model":"Gemini 3.8 Flash","vendor":"Google","level":"high","costUsd":29.16,"basis":"measured"},"claude-code":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":38.33,"basis":"rebuilt","taskScoreEstimated":true},"codex":{"model":"GPT-5.6 Sol","vendor":"OpenAI","level":"high","costUsd":28.95,"basis":"measured"},"antigravity":{"model":"Gemini 3.8 Flash","vendor":"Google","level":"high","costUsd":29.16,"basis":"measured"},"copilot":{"model":"GPT-5.6 Sol","vendor":"OpenAI","level":"high","costUsd":28.95,"basis":"measured"}}},"infra":{"pick":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":4.48,"basis":"rebuilt"},"runnerUp":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"medium","costUsd":4.97,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":4.48,"basis":"rebuilt"},"claude-code":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":4.48,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"medium","costUsd":4.97,"basis":"rebuilt"},"antigravity":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":4.48,"basis":"rebuilt"},"copilot":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":4.48,"basis":"rebuilt"}}},"algorithms":{"pick":{"model":"Muse Spark 1.1","vendor":"Meta","level":"xhigh","costUsd":23.34,"basis":"measured"},"runnerUp":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":24.02,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":24.62,"basis":"rebuilt"},"claude-code":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":24.62,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":24.02,"basis":"rebuilt"},"antigravity":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":24.62,"basis":"rebuilt"},"copilot":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":24.02,"basis":"rebuilt"}}},"ui":{"pick":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.33,"basis":"rebuilt"},"runnerUp":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.38,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.33,"basis":"rebuilt"},"claude-code":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.33,"basis":"rebuilt"},"codex":{"model":"GPT-6.1 Sol","vendor":"OpenAI","level":"high","costUsd":20.12,"basis":"rebuilt"},"antigravity":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.33,"basis":"rebuilt"},"copilot":{"model":"Claude Opus 5.5","vendor":"Anthropic","level":"xhigh","costUsd":18.33,"basis":"rebuilt"}}},"security":{"pick":{"model":"GPT-6 Sol","vendor":"OpenAI","level":"max","costUsd":17.65,"basis":"rebuilt"},"runnerUp":{"model":"MiMo-V2.6-Pro","vendor":"Xiaomi","level":"standard","costUsd":19.75,"basis":"rebuilt"},"byTool":{"cursor":{"model":"Grok 4.7","vendor":"SpaceXAI","level":"high","costUsd":20.6,"basis":"rebuilt"},"claude-code":{"model":"Claude Haiku 5.5","vendor":"Anthropic","level":"max","costUsd":20.82,"basis":"rebuilt"},"codex":{"model":"GPT-6 Sol","vendor":"OpenAI","level":"max","costUsd":17.65,"basis":"rebuilt"},"antigravity":{"model":"Claude Sonnet 5.5","vendor":"Anthropic","level":"xhigh","costUsd":23.48,"basis":"rebuilt"},"copilot":{"model":"GPT-6 Sol","vendor":"OpenAI","level":"max","costUsd":17.65,"basis":"rebuilt"}}}},"byPlan":{"claude-pro":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":34.29,"taskScoreEstimated":true},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Haiku 5.5","level":"max","costUsd":20.52}}},"claude-max":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Claude Fable 5.1","level":"high","costUsd":31.59},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Haiku 5.5","level":"max","costUsd":20.52}}},"claude-team":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":34.29,"taskScoreEstimated":true},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Haiku 5.5","level":"max","costUsd":20.52}}},"claude-team-premium":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Claude Fable 5.1","level":"high","costUsd":31.59},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Haiku 5.5","level":"max","costUsd":20.52}}},"claude-enterprise":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":38.33,"taskScoreEstimated":true},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":24.62},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"Claude Haiku 5.5","level":"max","costUsd":20.82}}},"chatgpt-plus":{"overall":{"model":"GPT-6.1 Sol","level":"medium","costUsd":9.58},"byTask":{"greenfield":{"model":"GPT-6.1 Sol","level":"high","costUsd":10.78},"legacy":{"model":"GPT-6 Astra","level":"low","costUsd":16.92},"explore":{"model":"GPT-6.1 Sol","level":"high","costUsd":22.43},"debug":{"model":"GPT-6.1 Sol","level":"medium","costUsd":9.96},"prod":{"model":"GPT-5.6 Sol","level":"medium","costUsd":24.15},"infra":{"model":"GPT-6.1 Sol","level":"medium","costUsd":3.55},"algorithms":{"model":"GPT-6.1 Sol","level":"medium","costUsd":22.17},"ui":{"model":"GPT-6.1 Sol","level":"high","costUsd":17.49},"security":{"model":"GPT-6 Sol","level":"xhigh","costUsd":15.45}}},"chatgpt-pro":{"overall":{"model":"GPT-6.1 Sol","level":"medium","costUsd":9.58},"byTask":{"greenfield":{"model":"GPT-6.1 Sol","level":"high","costUsd":10.78},"legacy":{"model":"GPT-6 Astra","level":"low","costUsd":16.92},"explore":{"model":"GPT-6.1 Sol","level":"high","costUsd":22.43},"debug":{"model":"GPT-6.1 Sol","level":"medium","costUsd":9.96},"prod":{"model":"GPT-5.6 Sol","level":"medium","costUsd":24.15},"infra":{"model":"GPT-6.1 Sol","level":"medium","costUsd":3.55},"algorithms":{"model":"GPT-6.1 Sol","level":"medium","costUsd":22.17},"ui":{"model":"GPT-6.1 Sol","level":"high","costUsd":17.49},"security":{"model":"GPT-6 Sol","level":"xhigh","costUsd":15.45}}},"chatgpt-enterprise":{"overall":{"model":"GPT-6.1 Sol","level":"high","costUsd":11.17},"byTask":{"greenfield":{"model":"GPT-6.1 Sol","level":"high","costUsd":13.05},"legacy":{"model":"GPT-6.1 Sol","level":"high","costUsd":22.2},"explore":{"model":"GPT-6.1 Sol","level":"high","costUsd":25.33},"debug":{"model":"GPT-6.1 Sol","level":"high","costUsd":11.57},"prod":{"model":"GPT-5.6 Sol","level":"high","costUsd":28.95},"infra":{"model":"GPT-6.1 Sol","level":"medium","costUsd":4.97},"algorithms":{"model":"GPT-6.1 Sol","level":"high","costUsd":24.02},"ui":{"model":"GPT-6.1 Sol","level":"high","costUsd":20.12},"security":{"model":"GPT-6 Sol","level":"max","costUsd":17.65}}},"google-ai-pro":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Gemini 3.8 Flash","level":"medium","costUsd":30.58},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":20.31}}},"google-ai-ultra":{"overall":{"model":"Claude Opus 5.5","level":"high","costUsd":8.04},"byTask":{"greenfield":{"model":"Claude Opus 5.5","level":"high","costUsd":9.2},"legacy":{"model":"Claude Opus 5.5","level":"high","costUsd":14.57},"explore":{"model":"Claude Opus 5.5","level":"high","costUsd":17.94},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":8.43},"prod":{"model":"Gemini 3.8 Flash","level":"medium","costUsd":30.58},"infra":{"model":"Claude Opus 5.5","level":"high","costUsd":2.09},"algorithms":{"model":"Claude Opus 5.5","level":"high","costUsd":20.12},"ui":{"model":"Claude Opus 5.5","level":"high","costUsd":14.19},"security":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":20.31}}},"gemini-enterprise":{"overall":{"model":"Gemini 3.8 Flash","level":"high","costUsd":10.43},"byTask":{"greenfield":{"model":"Gemini 3.8 Flash","level":"high","costUsd":16.23},"legacy":{"model":"Gemini 3.7 Flash","level":"high","costUsd":50.97},"explore":{"model":"Gemini 3.8 Flash","level":"high","costUsd":36.19},"debug":{"model":"Gemini 3.8 Flash","level":"high","costUsd":15.62},"prod":{"model":"Gemini 3.8 Flash","level":"high","costUsd":29.16},"infra":{"model":"Gemini 3.8 Flash","level":"high","costUsd":6.01},"algorithms":{"model":"Gemini 3.7 Flash","level":"high","costUsd":26.17},"ui":{"model":"Gemini 3.8 Flash","level":"high","costUsd":35.78},"security":{"model":"Gemini 3.7 Flash","level":"high","costUsd":29.4}}},"cursor":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"Gemini 3.8 Flash","level":"high","costUsd":29.16},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":24.62},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"Grok 4.7","level":"high","costUsd":20.6}}},"cursor-teams":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"Gemini 3.8 Flash","level":"high","costUsd":29.16},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":24.62},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"Grok 4.7","level":"high","costUsd":20.6}}},"cursor-enterprise":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"Gemini 3.8 Flash","level":"high","costUsd":29.16},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":24.62},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"Grok 4.7","level":"high","costUsd":20.6}}},"copilot-pro":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":18.91},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"Gemini 3.8 Flash","level":"high","costUsd":29.16},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":25.12},"ui":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":18.38},"security":{"model":"Grok 4.7","level":"high","costUsd":20.6}}},"copilot-pro-plus":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"GPT-5.6 Sol","level":"high","costUsd":28.95},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"GPT-6.1 Sol","level":"high","costUsd":24.02},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"GPT-6 Sol","level":"max","costUsd":17.65}}},"copilot-business":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"GPT-5.6 Sol","level":"high","costUsd":28.95},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"GPT-6.1 Sol","level":"high","costUsd":24.02},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"GPT-6 Sol","level":"max","costUsd":17.65}}},"copilot-enterprise":{"overall":{"model":"Claude Haiku 5.5","level":"max","costUsd":10.35},"byTask":{"greenfield":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":11.85},"legacy":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.74},"explore":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":21.17},"debug":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":10.87},"prod":{"model":"GPT-5.6 Sol","level":"high","costUsd":28.95},"infra":{"model":"Claude Sonnet 5.5","level":"xhigh","costUsd":4.48},"algorithms":{"model":"GPT-6.1 Sol","level":"high","costUsd":24.02},"ui":{"model":"Claude Opus 5.5","level":"xhigh","costUsd":18.33},"security":{"model":"GPT-6 Sol","level":"max","costUsd":17.65}}}}},"models":[{"name":"MiMo-V2.6-Pro","vendor":"Xiaomi","releaseDate":"2026-09-21","basis":"rebuilt","cursor":null,"openrouter":"xiaomi/mimo-v2.6-pro","plans":[],"overall":{"rank":1,"costUsd":9.23,"tokensUsd":0.23,"fixTimeUsd":9,"codingScore":77.5,"level":"standard","pricePerMTokUsd":0.6525},"byTask":{"greenfield":{"rank":3,"costUsd":12.75,"score":71.26,"level":"standard","status":"partial"},"legacy":{"rank":16,"costUsd":41.49,"score":43.01,"level":"standard","status":"partial"},"explore":{"rank":7,"costUsd":23.9,"score":56.79,"level":"standard","status":"estimated"},"debug":{"rank":5,"costUsd":13.99,"score":69.3,"level":"standard","status":"partial"},"prod":{"rank":7,"costUsd":36.14,"score":46.43,"level":"standard","status":"estimated"},"infra":{"rank":7,"costUsd":5.36,"score":85.74,"level":"standard","status":"partial"},"algorithms":{"rank":6,"costUsd":25.51,"score":55.17,"level":"standard"},"ui":{"rank":8,"costUsd":27.17,"score":53.6,"level":"standard"},"security":{"rank":3,"costUsd":19.75,"score":61.44,"level":"standard"}},"levels":[{"level":"standard","codingScore":77.5,"pricePerMTokUsd":0.6525,"thinkingMultiplier":1,"costUsd":9.23}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 34.9%, SciCode 60.9%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Gemini 4 Argon","vendor":"Google","releaseDate":"2026-09-30","basis":"rebuilt","cursor":null,"openrouter":null,"plans":[],"overall":{"rank":2,"costUsd":9.83,"tokensUsd":2.03,"fixTimeUsd":7.8,"codingScore":79.9,"level":"high","pricePerMTokUsd":6},"byTask":{"greenfield":{"rank":1,"costUsd":11.57,"score":76.62,"level":"high","status":"partial"},"legacy":{"rank":1,"costUsd":16.85,"score":68.17,"level":"high","status":"partial"},"explore":{"rank":16,"costUsd":29.31,"score":54.09,"level":"high"},"debug":{"rank":1,"costUsd":10.15,"score":79.27,"level":"high","status":"partial"},"prod":{"rank":8,"costUsd":36.41,"score":48.39,"level":"high","status":"estimated"},"infra":{"rank":2,"costUsd":4.7,"score":91.37,"level":"high","status":"partial"},"algorithms":{"rank":1,"costUsd":22.45,"score":61.03,"level":"high","status":"partial"},"ui":{"rank":4,"costUsd":23.24,"score":60.13,"level":"high"},"security":{"rank":1,"costUsd":17.55,"score":67.18,"level":"high","status":"partial"}},"levels":[{"level":"high","codingScore":79.9,"pricePerMTokUsd":6,"thinkingMultiplier":1,"costUsd":9.83}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 57.1%, SciCode 61.8%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Claude Haiku 5.5","vendor":"Anthropic","releaseDate":"2026-10-07","basis":"rebuilt","cursor":"Claude Haiku 5.5","openrouter":"anthropic/claude-haiku-5.5","plans":["claude-pro","claude-max","claude-team","claude-team-premium","claude-enterprise","cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":3,"costUsd":10.35,"tokensUsd":0.23,"fixTimeUsd":10.11,"codingScore":75.4,"level":"max","pricePerMTokUsd":0.3},"byTask":{"greenfield":{"rank":9,"costUsd":14.37,"score":68.71,"level":"max","status":"partial"},"legacy":{"rank":13,"costUsd":35.74,"score":46.71,"level":"max","status":"partial"},"explore":{"rank":19,"costUsd":31.44,"score":49.93,"level":"max"},"debug":{"rank":15,"costUsd":17.84,"score":63.82,"level":"max","status":"partial"},"prod":{"rank":22,"costUsd":43.24,"score":41.99,"level":"max","status":"estimated"},"infra":{"rank":4,"costUsd":5,"score":86.61,"level":"max","status":"partial"},"algorithms":{"rank":11,"costUsd":26.1,"score":54.6,"level":"max","status":"partial"},"ui":{"rank":10,"costUsd":27.86,"score":52.97,"level":"max"},"security":{"rank":6,"costUsd":20.82,"score":60.17,"level":"max","status":"partial"}},"levels":[{"level":"low","codingScore":69.8,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.39,"costUsd":13.5},{"level":"medium","codingScore":70.5,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.53,"costUsd":13.06},{"level":"high","codingScore":71.8,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.8,"costUsd":12.28},{"level":"xhigh","codingScore":73.9,"pricePerMTokUsd":0.3,"thinkingMultiplier":1.37,"costUsd":11.07},{"level":"max","codingScore":75.4,"pricePerMTokUsd":0.3,"thinkingMultiplier":3.9,"costUsd":10.35}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 32.8%, SciCode 55%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Gemini 3.8 Flash","vendor":"Google","releaseDate":"2026-09-02","basis":"measured","cursor":"Gemini 3.8 Flash","openrouter":"google/gemini-3.8-flash","plans":["google-ai-pro","google-ai-ultra","gemini-enterprise","cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":4,"costUsd":10.43,"tokensUsd":0.8,"fixTimeUsd":9.63,"codingScore":76.3,"level":"high","pricePerMTokUsd":2.25},"byTask":{"greenfield":{"rank":14,"costUsd":16.23,"score":66.92,"level":"high"},"legacy":{"rank":30,"costUsd":53.58,"score":37.37,"level":"high"},"explore":{"rank":29,"costUsd":36.19,"score":47.04,"level":"high"},"debug":{"rank":12,"costUsd":15.62,"score":67.8,"level":"high"},"prod":{"rank":2,"costUsd":29.16,"score":52.54,"level":"high"},"infra":{"rank":12,"costUsd":6.01,"score":85.39,"level":"high"},"algorithms":{"rank":15,"costUsd":26.47,"score":55,"level":"high"},"ui":{"rank":23,"costUsd":35.78,"score":47.33,"level":"high"},"security":{"rank":32,"costUsd":31.16,"score":50.85,"level":"high"}},"levels":[{"level":"low","codingScore":73.5,"pricePerMTokUsd":2.25,"thinkingMultiplier":0.35,"costUsd":11.79},{"level":"medium","codingScore":74.1,"pricePerMTokUsd":2.25,"thinkingMultiplier":0.6,"costUsd":11.52},{"level":"high","codingScore":76.3,"pricePerMTokUsd":2.25,"thinkingMultiplier":1,"costUsd":10.43}],"note":null},{"name":"Ling 3.1 Flash","vendor":"InclusionAI","releaseDate":"2026-10-01","basis":"rebuilt","cursor":null,"openrouter":"inclusionai/ling-3.1-flash","plans":[],"overall":{"rank":5,"costUsd":10.44,"tokensUsd":0.22,"fixTimeUsd":10.22,"codingScore":75.2,"level":"standard","pricePerMTokUsd":0.6},"byTask":{"greenfield":{"rank":12,"costUsd":15.61,"score":66.85,"level":"standard","status":"estimated"},"legacy":{"rank":12,"costUsd":35.48,"score":46.87,"level":"standard","status":"estimated"},"explore":{"rank":14,"costUsd":28.02,"score":52.8,"level":"standard","status":"estimated"},"debug":{"rank":11,"costUsd":15.27,"score":67.36,"level":"standard","status":"partial"},"prod":{"rank":11,"costUsd":37.72,"score":45.35,"level":"standard","status":"estimated"},"infra":{"rank":6,"costUsd":5.22,"score":86.03,"level":"standard","status":"partial"},"algorithms":{"rank":16,"costUsd":26.6,"score":54.1,"level":"standard","status":"partial"},"ui":{"rank":17,"costUsd":30.36,"score":50.78,"level":"standard","status":"estimated"},"security":{"rank":13,"costUsd":23.17,"score":57.53,"level":"standard","status":"estimated"}},"levels":[{"level":"standard","codingScore":75.2,"pricePerMTokUsd":0.6,"thinkingMultiplier":1,"costUsd":10.44}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 33.3%, SciCode 54.1%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Gemini 3.7 Flash","vendor":"Google","releaseDate":"2026-08-13","basis":"measured","cursor":"Gemini 3.7 Flash","openrouter":"google/gemini-3.7-flash","plans":["google-ai-pro","google-ai-ultra","gemini-enterprise","cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":6,"costUsd":10.53,"tokensUsd":0.8,"fixTimeUsd":9.74,"codingScore":76.1,"level":"high","pricePerMTokUsd":2.25},"byTask":{"greenfield":{"rank":23,"costUsd":19.26,"score":62.89,"level":"high","status":"partial"},"legacy":{"rank":26,"costUsd":50.97,"score":38.56,"level":"high","status":"partial"},"explore":{"rank":13,"costUsd":27.14,"score":54.36,"level":"high","status":"estimated"},"debug":{"rank":27,"costUsd":20.52,"score":61.35,"level":"high","status":"partial"},"prod":{"rank":12,"costUsd":37.82,"score":45.93,"level":"high","status":"estimated"},"infra":{"rank":18,"costUsd":6.76,"score":83.7,"level":"high"},"algorithms":{"rank":12,"costUsd":26.17,"score":55.29,"level":"high"},"ui":{"rank":26,"costUsd":40.63,"score":44.13,"level":"high"},"security":{"rank":30,"costUsd":29.4,"score":52.33,"level":"high","status":"partial"}},"levels":[{"level":"low","codingScore":71,"pricePerMTokUsd":2.25,"thinkingMultiplier":0.35,"costUsd":13.3},{"level":"medium","codingScore":71.5,"pricePerMTokUsd":2.25,"thinkingMultiplier":0.6,"costUsd":13.07},{"level":"high","codingScore":76.1,"pricePerMTokUsd":2.25,"thinkingMultiplier":1,"costUsd":10.53}],"note":null},{"name":"Grok 4.6","vendor":"SpaceXAI","releaseDate":"2026-08-12","basis":"measured","cursor":"Grok 4.6","openrouter":"x-ai/grok-4.6","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":7,"costUsd":10.77,"tokensUsd":1.41,"fixTimeUsd":9.36,"codingScore":76.8,"level":"high","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":18,"costUsd":17.75,"score":65.81,"level":"high"},"legacy":{"rank":10,"costUsd":35.04,"score":48.58,"level":"high","status":"partial"},"explore":{"rank":12,"costUsd":26.59,"score":55.7,"level":"high"},"debug":{"rank":22,"costUsd":19.45,"score":63.59,"level":"high","status":"partial"},"prod":{"rank":14,"costUsd":38.17,"score":46.38,"level":"high","status":"estimated"},"infra":{"rank":15,"costUsd":6.5,"score":85.55,"level":"high"},"algorithms":{"rank":18,"costUsd":27.16,"score":55.16,"level":"high"},"ui":{"rank":20,"costUsd":34.95,"score":48.64,"level":"high"},"security":{"rank":11,"costUsd":22.64,"score":59.81,"level":"high","status":"partial"}},"levels":[{"level":"low","codingScore":66.3,"pricePerMTokUsd":4,"thinkingMultiplier":0.35,"costUsd":16.96,"status":"below-quality-floor"},{"level":"medium","codingScore":74.4,"pricePerMTokUsd":4,"thinkingMultiplier":0.6,"costUsd":11.89},{"level":"high","codingScore":76.8,"pricePerMTokUsd":4,"thinkingMultiplier":1,"costUsd":10.77},{"level":"xhigh","codingScore":75.9,"pricePerMTokUsd":4,"thinkingMultiplier":1.7,"costUsd":11.66}],"note":null},{"name":"Muse Spark 1.3","vendor":"Meta","releaseDate":"2026-09-02","basis":"measured","cursor":"Muse Spark 1.3","openrouter":"meta/muse-spark-1.3","plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":8,"costUsd":10.89,"tokensUsd":1.37,"fixTimeUsd":9.52,"codingScore":76.5,"level":"xhigh","pricePerMTokUsd":2.75},"byTask":{"greenfield":{"rank":7,"costUsd":14.14,"score":71,"level":"xhigh","status":"partial"},"legacy":{"rank":14,"costUsd":36.6,"score":47.41,"level":"xhigh","status":"partial"},"explore":{"rank":11,"costUsd":25.59,"score":56.63,"level":"xhigh"},"debug":{"rank":10,"costUsd":15.25,"score":69.29,"level":"xhigh","status":"partial"},"prod":{"rank":40,"costUsd":65.42,"score":33.24,"level":"xhigh"},"infra":{"rank":19,"costUsd":6.76,"score":84.86,"level":"xhigh"},"algorithms":{"rank":7,"costUsd":25.65,"score":56.57,"level":"xhigh","status":"partial"},"ui":{"rank":7,"costUsd":26.34,"score":55.89,"level":"xhigh"},"security":{"rank":20,"costUsd":24.94,"score":57.29,"level":"xhigh"}},"levels":[{"level":"xhigh","codingScore":76.5,"pricePerMTokUsd":2.75,"thinkingMultiplier":2.03,"costUsd":10.89},{"level":"max","codingScore":75.8,"pricePerMTokUsd":2.75,"thinkingMultiplier":2.1,"costUsd":11.31}],"note":null},{"name":"Claude Sonnet 5.5","vendor":"Anthropic","releaseDate":"2026-09-28","basis":"rebuilt","cursor":"Claude Sonnet 5.5","openrouter":"anthropic/claude-sonnet-5.5","plans":["claude-pro","claude-max","claude-team","claude-team-premium","claude-enterprise","google-ai-pro","google-ai-ultra","cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":9,"costUsd":10.93,"tokensUsd":2.44,"fixTimeUsd":8.49,"codingScore":78.5,"level":"xhigh","pricePerMTokUsd":6},"byTask":{"greenfield":{"rank":2,"costUsd":11.85,"score":76.83,"level":"xhigh"},"legacy":{"rank":3,"costUsd":18.91,"score":65.95,"level":"xhigh","status":"partial"},"explore":{"rank":1,"costUsd":21.17,"score":63.1,"level":"xhigh"},"debug":{"rank":2,"costUsd":10.87,"score":78.63,"level":"xhigh","status":"partial"},"prod":{"rank":15,"costUsd":38.33,"score":47.48,"level":"xhigh","status":"estimated"},"infra":{"rank":1,"costUsd":4.48,"score":92.77,"level":"xhigh","status":"partial"},"algorithms":{"rank":5,"costUsd":25.12,"score":58.65,"level":"xhigh"},"ui":{"rank":2,"costUsd":18.38,"score":66.67,"level":"xhigh"},"security":{"rank":14,"costUsd":23.48,"score":60.42,"level":"xhigh","status":"partial"}},"levels":[{"level":"low","codingScore":71.7,"pricePerMTokUsd":6,"thinkingMultiplier":0.29,"costUsd":13.85},{"level":"medium","codingScore":74.3,"pricePerMTokUsd":6,"thinkingMultiplier":0.28,"costUsd":12.28},{"level":"high","codingScore":76.1,"pricePerMTokUsd":6,"thinkingMultiplier":1.03,"costUsd":11.89},{"level":"xhigh","codingScore":78.5,"pricePerMTokUsd":6,"thinkingMultiplier":1.46,"costUsd":10.93},{"level":"max","codingScore":80.2,"pricePerMTokUsd":6,"thinkingMultiplier":5,"costUsd":12.91}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 57.1%, SciCode 57.3%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Qwen3.8 Max (0902)","vendor":"Alibaba","releaseDate":"2026-09-02","basis":"measured","cursor":null,"openrouter":"qwen/qwen3.8-max-0902","plans":[],"overall":{"rank":10,"costUsd":11.1,"tokensUsd":1.42,"fixTimeUsd":9.68,"codingScore":76.2,"level":"standard","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":30,"costUsd":24,"score":58.32,"level":"standard","status":"partial"},"legacy":{"rank":39,"costUsd":102.89,"score":23.96,"level":"standard","status":"partial"},"explore":{"rank":6,"costUsd":23.43,"score":58.94,"level":"standard"},"debug":{"rank":26,"costUsd":20.06,"score":62.83,"level":"standard","status":"partial"},"prod":{"rank":34,"costUsd":48.7,"score":40.25,"level":"standard"},"infra":{"rank":10,"costUsd":5.63,"score":87.58,"level":"standard"},"algorithms":{"rank":23,"costUsd":27.61,"score":54.73,"level":"standard","status":"partial"},"ui":{"rank":25,"costUsd":36.84,"score":47.29,"level":"standard"},"security":{"rank":40,"costUsd":47.3,"score":40.97,"level":"standard","status":"partial"}},"levels":[{"level":"standard","codingScore":76.2,"pricePerMTokUsd":4,"thinkingMultiplier":1,"costUsd":11.1}],"note":null},{"name":"GPT-6.1 Sol","vendor":"OpenAI","releaseDate":"2026-09-29","basis":"rebuilt","cursor":null,"openrouter":"openai/gpt-6.1-sol","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":11,"costUsd":11.17,"tokensUsd":2.17,"fixTimeUsd":9,"codingScore":77.5,"level":"high","pricePerMTokUsd":6},"byTask":{"greenfield":{"rank":6,"costUsd":13.05,"score":74.2,"level":"high","status":"partial"},"legacy":{"rank":4,"costUsd":22.2,"score":61.43,"level":"high","status":"partial"},"explore":{"rank":10,"costUsd":25.33,"score":58.02,"level":"high"},"debug":{"rank":3,"costUsd":11.57,"score":76.78,"level":"high","status":"partial"},"prod":{"rank":5,"costUsd":34.92,"score":49.58,"level":"high","status":"estimated"},"infra":{"rank":3,"costUsd":4.97,"score":89.72,"level":"medium","status":"partial"},"algorithms":{"rank":3,"costUsd":24.02,"score":59.41,"level":"high","status":"partial"},"ui":{"rank":3,"costUsd":20.12,"score":63.94,"level":"high"},"security":{"rank":18,"costUsd":24.42,"score":58.98,"level":"high","status":"partial"}},"levels":[{"level":"low","codingScore":74.6,"pricePerMTokUsd":6,"thinkingMultiplier":0.27,"costUsd":12.09},{"level":"medium","codingScore":76.4,"pricePerMTokUsd":6,"thinkingMultiplier":0.46,"costUsd":11.24},{"level":"high","codingScore":77.5,"pricePerMTokUsd":6,"thinkingMultiplier":1.1,"costUsd":11.17},{"level":"xhigh","codingScore":77.7,"pricePerMTokUsd":6,"thinkingMultiplier":1.89,"costUsd":11.72},{"level":"max","codingScore":77.4,"pricePerMTokUsd":6,"thinkingMultiplier":3.42,"costUsd":13.17}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 51.5%, SciCode 55.8%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"DeepSeek V4.1 Flash","vendor":"DeepSeek","releaseDate":"2026-09-10","basis":"rebuilt","cursor":null,"openrouter":"deepseek/deepseek-v4.1-flash","plans":[],"overall":{"rank":12,"costUsd":11.41,"tokensUsd":0.29,"fixTimeUsd":11.12,"codingScore":73.6,"level":"max","pricePerMTokUsd":0.75},"byTask":{"greenfield":{"rank":4,"costUsd":12.97,"score":70.99,"level":"max","status":"partial"},"legacy":{"rank":15,"costUsd":37.42,"score":45.62,"level":"max","status":"partial"},"explore":{"rank":15,"costUsd":28.88,"score":52.13,"level":"max","status":"estimated"},"debug":{"rank":13,"costUsd":16.06,"score":66.33,"level":"max","status":"partial"},"prod":{"rank":6,"costUsd":35.52,"score":46.92,"level":"max"},"infra":{"rank":14,"costUsd":6.49,"score":83.25,"level":"max","status":"partial"},"algorithms":{"rank":25,"costUsd":28.02,"score":52.88,"level":"max"},"ui":{"rank":13,"costUsd":28.31,"score":52.63,"level":"max"},"security":{"rank":5,"costUsd":20.77,"score":60.3,"level":"max"}},"levels":[{"level":"non-reasoning","codingScore":62,"pricePerMTokUsd":0.75,"thinkingMultiplier":0.33,"costUsd":19.24,"status":"below-quality-floor"},{"level":"max","codingScore":73.6,"pricePerMTokUsd":0.75,"thinkingMultiplier":1.13,"costUsd":11.41}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 26.8%, SciCode 51.9%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"GLM-5.3","vendor":"Z AI","releaseDate":"2026-08-18","basis":"measured","cursor":"GLM 5.3","openrouter":"z-ai/glm-5.3","plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":13,"costUsd":11.46,"tokensUsd":1.02,"fixTimeUsd":10.44,"codingScore":74.8,"level":"max","pricePerMTokUsd":2.9},"byTask":{"greenfield":{"rank":15,"costUsd":16.25,"score":67.22,"level":"max"},"legacy":{"rank":24,"costUsd":49.9,"score":39.26,"level":"max","status":"partial"},"explore":{"rank":9,"costUsd":25.09,"score":56.63,"level":"max"},"debug":{"rank":6,"costUsd":14.57,"score":69.7,"level":"max","status":"partial"},"prod":{"rank":13,"costUsd":37.93,"score":46.08,"level":"max"},"infra":{"rank":13,"costUsd":6.09,"score":85.63,"level":"max"},"algorithms":{"rank":9,"costUsd":25.81,"score":55.91,"level":"max"},"ui":{"rank":18,"costUsd":32.69,"score":49.87,"level":"max"},"security":{"rank":19,"costUsd":24.59,"score":57.14,"level":"max"}},"levels":[{"level":"low","codingScore":71.4,"pricePerMTokUsd":2.9,"thinkingMultiplier":0.94,"costUsd":13.49},{"level":"max","codingScore":74.8,"pricePerMTokUsd":2.9,"thinkingMultiplier":0.93,"costUsd":11.46}],"note":null},{"name":"Qwen3.8-Flash-Next","vendor":"Alibaba","releaseDate":"2026-08-26","basis":"measured","cursor":null,"openrouter":null,"plans":[],"overall":{"rank":14,"costUsd":11.52,"tokensUsd":0.11,"fixTimeUsd":11.41,"codingScore":73.1,"level":"standard","pricePerMTokUsd":0.31},"byTask":{"greenfield":{"rank":27,"costUsd":21.95,"score":58.7,"level":"standard","status":"partial"},"legacy":{"rank":21,"costUsd":48.16,"score":39.27,"level":"standard","status":"estimated"},"explore":{"rank":23,"costUsd":32.23,"score":49.16,"level":"standard","status":"estimated"},"debug":{"rank":24,"costUsd":19.92,"score":61.04,"level":"standard","status":"partial"},"prod":{"rank":17,"costUsd":39.66,"score":43.99,"level":"standard","status":"estimated"},"infra":{"rank":8,"costUsd":5.47,"score":85.22,"level":"standard"},"algorithms":{"rank":34,"costUsd":30.43,"score":50.6,"level":"standard","status":"partial"},"ui":{"rank":14,"costUsd":29.15,"score":51.67,"level":"standard","status":"partial"},"security":{"rank":17,"costUsd":24.19,"score":56.32,"level":"standard","status":"estimated"}},"levels":[{"level":"standard","codingScore":73.1,"pricePerMTokUsd":0.31,"thinkingMultiplier":1,"costUsd":11.52}],"note":null},{"name":"Grok 4.7","vendor":"SpaceXAI","releaseDate":"2026-09-21","basis":"rebuilt","cursor":"Grok 4.7","openrouter":"x-ai/grok-4.7","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":15,"costUsd":11.57,"tokensUsd":1.34,"fixTimeUsd":10.22,"codingScore":75.2,"level":"high","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":8,"costUsd":14.17,"score":70.86,"level":"high"},"legacy":{"rank":11,"costUsd":35.39,"score":48.22,"level":"high","status":"partial"},"explore":{"rank":4,"costUsd":22.68,"score":59.63,"level":"high"},"debug":{"rank":7,"costUsd":14.67,"score":70.09,"level":"high","status":"partial"},"prod":{"rank":27,"costUsd":45.07,"score":42.08,"level":"high"},"infra":{"rank":17,"costUsd":6.57,"score":85.2,"level":"high","status":"partial"},"algorithms":{"rank":13,"costUsd":26.18,"score":55.99,"level":"high","status":"partial"},"ui":{"rank":9,"costUsd":27.45,"score":54.76,"level":"high"},"security":{"rank":4,"costUsd":20.6,"score":62.03,"level":"high"}},"levels":[{"level":"low","codingScore":72.6,"pricePerMTokUsd":4,"thinkingMultiplier":0.59,"costUsd":12.94},{"level":"high","codingScore":75.2,"pricePerMTokUsd":4,"thinkingMultiplier":0.84,"costUsd":11.57},{"level":"xhigh","codingScore":75.2,"pricePerMTokUsd":4,"thinkingMultiplier":1.2,"costUsd":11.77}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 24.8%, SciCode 57.8%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"MiMo-V2.6-Flash","vendor":"Xiaomi","releaseDate":"2026-09-21","basis":"rebuilt","cursor":null,"openrouter":"xiaomi/mimo-v2.6-flash","plans":[],"overall":{"rank":16,"costUsd":11.66,"tokensUsd":0.08,"fixTimeUsd":11.58,"codingScore":72.8,"level":"standard","pricePerMTokUsd":0.21},"byTask":{"greenfield":{"rank":13,"costUsd":15.94,"score":66.16,"level":"standard","status":"partial"},"legacy":{"rank":20,"costUsd":44.88,"score":40.93,"level":"standard","status":"partial"},"explore":{"rank":24,"costUsd":32.85,"score":48.64,"level":"standard","status":"estimated"},"debug":{"rank":18,"costUsd":18.14,"score":63.2,"level":"standard","status":"partial"},"prod":{"rank":18,"costUsd":39.91,"score":43.79,"level":"standard","status":"estimated"},"infra":{"rank":11,"costUsd":5.87,"score":84.23,"level":"standard","status":"partial"},"algorithms":{"rank":21,"costUsd":27.53,"score":53.06,"level":"standard","status":"partial"},"ui":{"rank":16,"costUsd":29.54,"score":51.3,"level":"standard"},"security":{"rank":7,"costUsd":20.86,"score":59.88,"level":"standard","status":"partial"}},"levels":[{"level":"standard","codingScore":72.8,"pricePerMTokUsd":0.21,"thinkingMultiplier":1,"costUsd":11.66}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 22.7%, SciCode 51.3%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Mistral Large 4 Preview","vendor":"Mistral","releaseDate":"2026-10-06","basis":"rebuilt","cursor":null,"openrouter":null,"plans":[],"overall":{"rank":17,"costUsd":11.73,"tokensUsd":1.01,"fixTimeUsd":10.72,"codingScore":74.3,"level":"standard","pricePerMTokUsd":2.77},"byTask":{"greenfield":{"rank":19,"costUsd":17.99,"score":64.81,"level":"standard","status":"estimated"},"legacy":{"rank":17,"costUsd":41.8,"score":43.61,"level":"standard","status":"estimated"},"explore":{"rank":18,"costUsd":30.96,"score":51.24,"level":"standard","status":"estimated"},"debug":{"rank":19,"costUsd":18.51,"score":64.13,"level":"standard","status":"partial"},"prod":{"rank":19,"costUsd":39.92,"score":44.76,"level":"standard","status":"estimated"},"infra":{"rank":16,"costUsd":6.51,"score":84.63,"level":"standard","status":"partial"},"algorithms":{"rank":22,"costUsd":27.58,"score":54.2,"level":"standard","status":"partial"},"ui":{"rank":21,"costUsd":35.09,"score":48.04,"level":"standard","status":"estimated"},"security":{"rank":22,"costUsd":25.84,"score":55.85,"level":"standard","status":"partial"}},"levels":[{"level":"standard","codingScore":74.3,"pricePerMTokUsd":2.77,"thinkingMultiplier":1,"costUsd":11.73}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 26.8%, SciCode 54.2%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"Claude Opus 5.5","vendor":"Anthropic","releaseDate":"2026-09-22","basis":"rebuilt","cursor":"Claude Opus 5.5","openrouter":"anthropic/claude-opus-5.5","plans":["claude-pro","claude-max","claude-team","claude-team-premium","claude-enterprise","google-ai-pro","google-ai-ultra","cursor","cursor-teams","cursor-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":18,"costUsd":11.77,"tokensUsd":3.38,"fixTimeUsd":8.39,"codingScore":78.7,"level":"medium","pricePerMTokUsd":12},"byTask":{"greenfield":{"rank":5,"costUsd":13.02,"score":78.91,"level":"xhigh"},"legacy":{"rank":2,"costUsd":18.74,"score":69.82,"level":"xhigh","status":"partial"},"explore":{"rank":2,"costUsd":22.32,"score":65.14,"level":"xhigh"},"debug":{"rank":4,"costUsd":12.31,"score":80.19,"level":"xhigh","status":"partial"},"prod":{"rank":38,"costUsd":54.27,"score":40.73,"level":"xhigh"},"infra":{"rank":5,"costUsd":5.19,"score":92.99,"level":"medium","status":"partial"},"algorithms":{"rank":4,"costUsd":24.62,"score":62.45,"level":"xhigh"},"ui":{"rank":1,"costUsd":18.33,"score":70.4,"level":"xhigh"},"security":{"rank":26,"costUsd":27.72,"score":59.15,"level":"xhigh"}},"levels":[{"level":"low","codingScore":76.4,"pricePerMTokUsd":12,"thinkingMultiplier":0.32,"costUsd":12.66},{"level":"medium","codingScore":78.7,"pricePerMTokUsd":12,"thinkingMultiplier":0.55,"costUsd":11.77},{"level":"high","codingScore":79.4,"pricePerMTokUsd":12,"thinkingMultiplier":0.81,"costUsd":11.81},{"level":"xhigh","codingScore":81.2,"pricePerMTokUsd":12,"thinkingMultiplier":1.38,"costUsd":11.77},{"level":"max","codingScore":81.8,"pricePerMTokUsd":12,"thinkingMultiplier":4.51,"costUsd":16.42}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 52.5%, SciCode 59.3%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"GPT-6 Sol","vendor":"OpenAI","releaseDate":"2026-09-22","basis":"rebuilt","cursor":null,"openrouter":"openai/gpt-6-sol","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":19,"costUsd":12.4,"tokensUsd":3.35,"fixTimeUsd":9.05,"codingScore":77.4,"level":"max","pricePerMTokUsd":6},"byTask":{"greenfield":{"rank":11,"costUsd":15.38,"score":72.43,"level":"max","status":"partial"},"legacy":{"rank":7,"costUsd":28.6,"score":56.36,"level":"max","status":"partial"},"explore":{"rank":17,"costUsd":30.05,"score":55.02,"level":"max"},"debug":{"rank":9,"costUsd":15.16,"score":72.77,"level":"max","status":"partial"},"prod":{"rank":9,"costUsd":36.94,"score":49.44,"level":"max"},"infra":{"rank":20,"costUsd":6.93,"score":88.55,"level":"max","status":"partial"},"algorithms":{"rank":14,"costUsd":26.46,"score":58.47,"level":"max"},"ui":{"rank":6,"costUsd":25.95,"score":58.99,"level":"max"},"security":{"rank":2,"costUsd":17.65,"score":69.05,"level":"max"}},"levels":[{"level":"non-reasoning","codingScore":69.4,"pricePerMTokUsd":6,"thinkingMultiplier":0.15,"costUsd":15.21,"status":"below-quality-floor"},{"level":"low","codingScore":68.8,"pricePerMTokUsd":6,"thinkingMultiplier":0.35,"costUsd":15.8,"status":"below-quality-floor"},{"level":"medium","codingScore":72.8,"pricePerMTokUsd":6,"thinkingMultiplier":0.6,"costUsd":13.45},{"level":"high","codingScore":74.5,"pricePerMTokUsd":6,"thinkingMultiplier":1,"costUsd":12.79},{"level":"xhigh","codingScore":75.1,"pricePerMTokUsd":6,"thinkingMultiplier":1.7,"costUsd":13.04},{"level":"max","codingScore":77.4,"pricePerMTokUsd":6,"thinkingMultiplier":2.5,"costUsd":12.4}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 43.9%, SciCode 57.6%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"GPT-6 Luna","vendor":"OpenAI","releaseDate":"2026-09-22","basis":"rebuilt","cursor":null,"openrouter":"openai/gpt-6-luna","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":20,"costUsd":12.47,"tokensUsd":0.17,"fixTimeUsd":12.3,"codingScore":71.6,"level":"max","pricePerMTokUsd":0.3},"byTask":{"greenfield":{"rank":10,"costUsd":14.47,"score":68.45,"level":"max","status":"partial"},"legacy":{"rank":19,"costUsd":44.77,"score":41.08,"level":"max","status":"partial"},"explore":{"rank":32,"costUsd":40.23,"score":43.69,"level":"max"},"debug":{"rank":17,"costUsd":18.1,"score":63.39,"level":"max","status":"partial"},"prod":{"rank":31,"costUsd":46.03,"score":40.4,"level":"max","status":"estimated"},"infra":{"rank":21,"costUsd":6.97,"score":81.96,"level":"max","status":"partial"},"algorithms":{"rank":8,"costUsd":25.78,"score":54.81,"level":"max"},"ui":{"rank":19,"costUsd":33.15,"score":48.52,"level":"max"},"security":{"rank":8,"costUsd":21.01,"score":59.84,"level":"max"}},"levels":[{"level":"non-reasoning","codingScore":46.9,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.24,"costUsd":35.22,"status":"below-quality-floor"},{"level":"low","codingScore":51.9,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.41,"costUsd":28.85,"status":"below-quality-floor"},{"level":"medium","codingScore":63,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.6,"costUsd":18.31,"status":"below-quality-floor"},{"level":"high","codingScore":65.9,"pricePerMTokUsd":0.3,"thinkingMultiplier":0.79,"costUsd":16.15,"status":"below-quality-floor"},{"level":"xhigh","codingScore":68.8,"pricePerMTokUsd":0.3,"thinkingMultiplier":1.3,"costUsd":14.19,"status":"below-quality-floor"},{"level":"max","codingScore":71.6,"pricePerMTokUsd":0.3,"thinkingMultiplier":2.28,"costUsd":12.47}],"note":"Artificial Analysis no longer publishes a Coding Index for new models, so it is rebuilt the way AA built it (67% Terminal-Bench, 32% SciCode) from AA's current results, both measured on this model at this effort level: Terminal-Bench 4.0 12.6%, SciCode 54.6%. Terminal-Bench 4.0 is put on the older 2.1 scale using the 62 models AA ran on both. Typical miss about ±2.6 points in back-tests on models with a published Coding Index. Replaced automatically if AA publishes one."},{"name":"GLM-5.3-Flash","vendor":"Z AI","releaseDate":"2026-08-26","basis":"measured","cursor":"GLM 5.3 Flash","openrouter":"z-ai/glm-5.3-flash","plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":21,"costUsd":12.48,"tokensUsd":0.12,"fixTimeUsd":12.36,"codingScore":71.5,"level":"standard","pricePerMTokUsd":0.325},"byTask":{"greenfield":{"rank":34,"costUsd":25.52,"score":55.01,"level":"standard"},"legacy":{"rank":40,"costUsd":144.07,"score":17.76,"level":"standard","status":"partial"},"explore":{"rank":28,"costUsd":36.01,"score":46.39,"level":"standard","status":"estimated"},"debug":{"rank":14,"costUsd":16.64,"score":65.25,"level":"standard","status":"partial"},"prod":{"rank":3,"costUsd":29.67,"score":51.24,"level":"standard"},"infra":{"rank":9,"costUsd":5.53,"score":85.1,"level":"standard"},"algorithms":{"rank":20,"costUsd":27.43,"score":53.21,"level":"standard"},"ui":{"rank":39,"costUsd":83.04,"score":27.26,"level":"standard"},"security":{"rank":12,"costUsd":22.68,"score":57.91,"level":"standard"}},"levels":[{"level":"standard","codingScore":71.5,"pricePerMTokUsd":0.325,"thinkingMultiplier":1,"costUsd":12.48}],"note":null},{"name":"Kimi K3","vendor":"Kimi","releaseDate":"2026-07-16","basis":"measured","cursor":"Kimi K3","openrouter":"moonshotai/kimi-k3","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":22,"costUsd":12.79,"tokensUsd":3.11,"fixTimeUsd":9.68,"codingScore":76.2,"level":"max","pricePerMTokUsd":9},"byTask":{"greenfield":{"rank":20,"costUsd":18.05,"score":68.04,"level":"max","status":"partial"},"legacy":{"rank":36,"costUsd":80.67,"score":29.88,"level":"max","status":"partial"},"explore":{"rank":3,"costUsd":22.47,"score":62.41,"level":"max"},"debug":{"rank":28,"costUsd":21.05,"score":64.11,"level":"max","status":"partial"},"prod":{"rank":16,"costUsd":38.98,"score":47.69,"level":"max"},"infra":{"rank":29,"costUsd":8.79,"score":83.87,"level":"max"},"algorithms":{"rank":29,"costUsd":29.02,"score":55.6,"level":"max"},"ui":{"rank":15,"costUsd":29.34,"score":55.3,"level":"max"},"security":{"rank":29,"costUsd":28.72,"score":55.88,"level":"max","status":"partial"}},"levels":[{"level":"low","codingScore":72,"pricePerMTokUsd":9,"thinkingMultiplier":0.93,"costUsd":15.34},{"level":"max","codingScore":76.2,"pricePerMTokUsd":9,"thinkingMultiplier":0.94,"costUsd":12.79}],"note":null},{"name":"GPT-5.6 Luna","vendor":"OpenAI","releaseDate":"2026-07-09","basis":"measured","cursor":null,"openrouter":"openai/gpt-5.6-luna","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":23,"costUsd":12.84,"tokensUsd":0.42,"fixTimeUsd":12.42,"codingScore":71.4,"level":"max","pricePerMTokUsd":0.7},"byTask":{"greenfield":{"rank":17,"costUsd":17.48,"score":64.57,"level":"max","status":"partial"},"legacy":{"rank":23,"costUsd":49.17,"score":39.04,"level":"max","status":"partial"},"explore":{"rank":26,"costUsd":35.04,"score":47.4,"level":"max"},"debug":{"rank":21,"costUsd":19.42,"score":62.08,"level":"max","status":"partial"},"prod":{"rank":32,"costUsd":46.63,"score":40.32,"level":"max"},"infra":{"rank":23,"costUsd":7.54,"score":81.21,"level":"max"},"algorithms":{"rank":10,"costUsd":26.03,"score":54.89,"level":"max"},"ui":{"rank":29,"costUsd":43.26,"score":42.15,"level":"max"},"security":{"rank":10,"costUsd":21.97,"score":59.1,"level":"max","status":"partial"}},"levels":[{"level":"non-reasoning","codingScore":39.3,"pricePerMTokUsd":0.7,"thinkingMultiplier":0.15,"costUsd":48.2,"status":"below-quality-floor"},{"level":"low","codingScore":44.2,"pricePerMTokUsd":0.7,"thinkingMultiplier":0.35,"costUsd":39.45,"status":"below-quality-floor"},{"level":"medium","codingScore":50.7,"pricePerMTokUsd":0.7,"thinkingMultiplier":0.6,"costUsd":30.46,"status":"below-quality-floor"},{"level":"high","codingScore":63.3,"pricePerMTokUsd":0.7,"thinkingMultiplier":1,"costUsd":18.27,"status":"below-quality-floor"},{"level":"xhigh","codingScore":68.6,"pricePerMTokUsd":0.7,"thinkingMultiplier":1.7,"costUsd":14.54,"status":"below-quality-floor"},{"level":"max","codingScore":71.4,"pricePerMTokUsd":0.7,"thinkingMultiplier":2.5,"costUsd":12.84}],"note":null},{"name":"GPT-5.6 Sol","vendor":"OpenAI","releaseDate":"2026-07-09","basis":"measured","cursor":null,"openrouter":"openai/gpt-5.6-sol","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":24,"costUsd":13.2,"tokensUsd":3.57,"fixTimeUsd":9.63,"codingScore":76.3,"level":"medium","pricePerMTokUsd":12},"byTask":{"greenfield":{"rank":21,"costUsd":18.19,"score":68.56,"level":"medium"},"legacy":{"rank":8,"costUsd":33.54,"score":53.05,"level":"high","status":"partial"},"explore":{"rank":33,"costUsd":42.01,"score":46.9,"level":"high"},"debug":{"rank":20,"costUsd":18.7,"score":67.84,"level":"medium"},"prod":{"rank":1,"costUsd":28.95,"score":57.11,"level":"high"},"infra":{"rank":24,"costUsd":7.92,"score":86.65,"level":"medium"},"algorithms":{"rank":17,"costUsd":26.91,"score":59.13,"level":"high"},"ui":{"rank":22,"costUsd":35.4,"score":51.57,"level":"high"},"security":{"rank":9,"costUsd":21.27,"score":64.51,"level":"medium","status":"partial"}},"levels":[{"level":"non-reasoning","codingScore":65.1,"pricePerMTokUsd":12,"thinkingMultiplier":0.15,"costUsd":19.9,"status":"below-quality-floor"},{"level":"low","codingScore":69.7,"pricePerMTokUsd":12,"thinkingMultiplier":0.35,"costUsd":16.92},{"level":"medium","codingScore":76.3,"pricePerMTokUsd":12,"thinkingMultiplier":0.6,"costUsd":13.2},{"level":"high","codingScore":77.2,"pricePerMTokUsd":12,"thinkingMultiplier":1,"costUsd":13.35},{"level":"xhigh","codingScore":78.3,"pricePerMTokUsd":12,"thinkingMultiplier":1.7,"costUsd":13.89},{"level":"max","codingScore":77.4,"pricePerMTokUsd":12,"thinkingMultiplier":2.5,"costUsd":15.75}],"note":null},{"name":"Muse Spark 1.2","vendor":"Meta","releaseDate":"2026-08-05","basis":"measured","cursor":null,"openrouter":"meta/muse-spark-1.2","plans":[],"overall":{"rank":25,"costUsd":13.25,"tokensUsd":1.32,"fixTimeUsd":11.94,"codingScore":72.2,"level":"xhigh","pricePerMTokUsd":2.75},"byTask":{"greenfield":{"rank":26,"costUsd":21.87,"score":60.43,"level":"xhigh","status":"partial"},"legacy":{"rank":35,"costUsd":75.68,"score":29.95,"level":"xhigh","status":"partial"},"explore":{"rank":35,"costUsd":44,"score":42.6,"level":"xhigh","status":"estimated"},"debug":{"rank":34,"costUsd":27.58,"score":54.54,"level":"xhigh","status":"partial"},"prod":{"rank":20,"costUsd":42.61,"score":43.41,"level":"xhigh","status":"estimated"},"infra":{"rank":30,"costUsd":8.82,"score":80.24,"level":"xhigh"},"algorithms":{"rank":26,"costUsd":28.25,"score":53.92,"level":"xhigh","status":"partial"},"ui":{"rank":27,"costUsd":41.62,"score":44,"level":"xhigh"},"security":{"rank":16,"costUsd":23.99,"score":58.11,"level":"xhigh","status":"partial"}},"levels":[{"level":"xhigh","codingScore":72.2,"pricePerMTokUsd":2.75,"thinkingMultiplier":1.7,"costUsd":13.25}],"note":null},{"name":"Grok 4.5","vendor":"SpaceXAI","releaseDate":"2026-07-08","basis":"measured","cursor":"Grok 4.5","openrouter":"x-ai/grok-4.5","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":26,"costUsd":13.31,"tokensUsd":1.49,"fixTimeUsd":11.82,"codingScore":72.4,"level":"high","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":35,"costUsd":25.88,"score":56.4,"level":"high","status":"partial"},"legacy":{"rank":29,"costUsd":53.15,"score":38.12,"level":"high","status":"partial"},"explore":{"rank":27,"costUsd":35.9,"score":47.95,"level":"high","status":"estimated"},"debug":{"rank":35,"costUsd":27.74,"score":54.61,"level":"high","status":"partial"},"prod":{"rank":4,"costUsd":32.84,"score":50.25,"level":"high","status":"estimated"},"infra":{"rank":28,"costUsd":8.48,"score":81.26,"level":"high"},"algorithms":{"rank":28,"costUsd":28.54,"score":53.88,"level":"high","status":"partial"},"ui":{"rank":31,"costUsd":47.71,"score":40.76,"level":"high"},"security":{"rank":15,"costUsd":23.87,"score":58.47,"level":"high","status":"partial"}},"levels":[{"level":"high","codingScore":72.4,"pricePerMTokUsd":4,"thinkingMultiplier":1,"costUsd":13.31}],"note":null},{"name":"Qwen3.8 2.4T A95B","vendor":"Alibaba","releaseDate":"2026-08-12","basis":"measured","cursor":null,"openrouter":"qwen/qwen3.8-2.4t-a95b","plans":[],"overall":{"rank":27,"costUsd":13.62,"tokensUsd":1.5,"fixTimeUsd":12.12,"codingScore":71.9,"level":"standard","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":28,"costUsd":23.04,"score":59.37,"level":"standard","status":"estimated"},"legacy":{"rank":31,"costUsd":60.87,"score":34.92,"level":"standard","status":"estimated"},"explore":{"rank":30,"costUsd":37.13,"score":47.08,"level":"standard","status":"estimated"},"debug":{"rank":30,"costUsd":25.85,"score":56.43,"level":"standard","status":"partial"},"prod":{"rank":21,"costUsd":43.24,"score":43.21,"level":"standard","status":"estimated"},"infra":{"rank":25,"costUsd":8.29,"score":81.66,"level":"standard"},"algorithms":{"rank":27,"costUsd":28.3,"score":54.1,"level":"standard","status":"partial"},"ui":{"rank":32,"costUsd":47.78,"score":40.72,"level":"standard","status":"estimated"},"security":{"rank":41,"costUsd":51.53,"score":38.87,"level":"standard","status":"partial"}},"levels":[{"level":"standard","codingScore":71.9,"pricePerMTokUsd":4,"thinkingMultiplier":1,"costUsd":13.62}],"note":null},{"name":"Composer 2.5","vendor":"Cursor","releaseDate":"2026-05-18","basis":"hand-estimated","cursor":"Composer 2.5","openrouter":null,"plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":28,"costUsd":13.67,"tokensUsd":0.58,"fixTimeUsd":13.1,"codingScore":70.3,"level":"standard","pricePerMTokUsd":1.5},"byTask":{"greenfield":{"rank":40,"costUsd":32.04,"score":49.82,"level":"standard","status":"partial"},"legacy":{"rank":41,"costUsd":1539.25,"score":2,"level":"standard","status":"partial"},"explore":{"rank":31,"costUsd":39.87,"score":44.31,"level":"standard","status":"estimated"},"debug":{"rank":32,"costUsd":27.42,"score":53.75,"level":"standard","status":"estimated"},"prod":{"rank":23,"costUsd":43.46,"score":42.18,"level":"standard","status":"estimated"},"infra":{"rank":27,"costUsd":8.37,"score":79.77,"level":"standard","status":"estimated"},"algorithms":{"rank":24,"costUsd":27.9,"score":53.32,"level":"standard","status":"estimated"},"ui":{"rank":41,"costUsd":102.87,"score":23.46,"level":"standard","status":"partial"},"security":{"rank":21,"costUsd":25.81,"score":55.28,"level":"standard","status":"estimated"}},"levels":[{"level":"standard","codingScore":70.3,"pricePerMTokUsd":1.5,"thinkingMultiplier":1,"costUsd":13.67}],"note":"Not in the Artificial Analysis API. Artificial Analysis's Coding Agent Index (May 2026) scores Composer 2.5 at 62, vs 66 for Claude Opus 4.7 (max) and 65 for GPT-5.5 (xhigh). Those two score 73.6 and 74.9 on the Coding Index, about 1.13× their agent-index score; 62 × 1.13 ≈ 70.3 (range 69–71). Price is Cursor's standard tier; the \"Fast\" tier costs $3 / $15."},{"name":"Qwen3.8 Max","vendor":"Alibaba","releaseDate":"2026-08-03","basis":"measured","cursor":null,"openrouter":null,"plans":[],"overall":{"rank":29,"costUsd":13.68,"tokensUsd":1.5,"fixTimeUsd":12.18,"codingScore":71.8,"level":"standard","pricePerMTokUsd":4},"byTask":{"greenfield":{"rank":29,"costUsd":24,"score":58.32,"level":"standard","status":"partial"},"legacy":{"rank":38,"costUsd":102.89,"score":23.96,"level":"standard","status":"partial"},"explore":{"rank":5,"costUsd":23.43,"score":58.94,"level":"standard"},"debug":{"rank":25,"costUsd":20.06,"score":62.83,"level":"standard","status":"partial"},"prod":{"rank":33,"costUsd":48.7,"score":40.25,"level":"standard"},"infra":{"rank":22,"costUsd":7.27,"score":83.83,"level":"standard"},"algorithms":{"rank":19,"costUsd":27.22,"score":55.1,"level":"standard","status":"partial"},"ui":{"rank":24,"costUsd":36.84,"score":47.29,"level":"standard"},"security":{"rank":39,"costUsd":47.3,"score":40.97,"level":"standard","status":"partial"}},"levels":[{"level":"standard","codingScore":71.8,"pricePerMTokUsd":4,"thinkingMultiplier":1,"costUsd":13.68}],"note":null},{"name":"Muse Spark 1.1","vendor":"Meta","releaseDate":"2026-07-09","basis":"measured","cursor":null,"openrouter":"meta/muse-spark-1.1","plans":[],"overall":{"rank":30,"costUsd":13.81,"tokensUsd":1.33,"fixTimeUsd":12.48,"codingScore":71.3,"level":"xhigh","pricePerMTokUsd":2.75},"byTask":{"greenfield":{"rank":33,"costUsd":24.75,"score":57.31,"level":"xhigh","status":"partial"},"legacy":{"rank":34,"costUsd":71.67,"score":31.12,"level":"xhigh","status":"partial"},"explore":{"rank":36,"costUsd":44.71,"score":42.2,"level":"xhigh"},"debug":{"rank":33,"costUsd":27.56,"score":54.56,"level":"xhigh"},"prod":{"rank":24,"costUsd":43.61,"score":42.82,"level":"xhigh","status":"estimated"},"infra":{"rank":31,"costUsd":9.41,"score":79.06,"level":"xhigh"},"algorithms":{"rank":2,"costUsd":23.34,"score":58.8,"level":"xhigh","status":"partial"},"ui":{"rank":30,"costUsd":46.05,"score":41.47,"level":"xhigh"},"security":{"rank":23,"costUsd":26.49,"score":55.57,"level":"xhigh","status":"estimated"}},"levels":[{"level":"xhigh","codingScore":71.3,"pricePerMTokUsd":2.75,"thinkingMultiplier":1.7,"costUsd":13.81}],"note":null},{"name":"GPT-5.6 Terra","vendor":"OpenAI","releaseDate":"2026-07-09","basis":"measured","cursor":null,"openrouter":"openai/gpt-5.6-terra","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":31,"costUsd":14.43,"tokensUsd":5.02,"fixTimeUsd":9.42,"codingScore":76.7,"level":"max","pricePerMTokUsd":7},"byTask":{"greenfield":{"rank":25,"costUsd":21.49,"score":66.39,"level":"max","status":"partial"},"legacy":{"rank":27,"costUsd":51.17,"score":42.41,"level":"max","status":"partial"},"explore":{"rank":20,"costUsd":31.9,"score":55.4,"level":"max","status":"estimated"},"debug":{"rank":29,"costUsd":21.77,"score":66.04,"level":"max","status":"partial"},"prod":{"rank":10,"costUsd":37.28,"score":51.04,"level":"max"},"infra":{"rank":32,"costUsd":9.54,"score":85.96,"level":"max"},"algorithms":{"rank":33,"costUsd":30.2,"score":56.94,"level":"max"},"ui":{"rank":35,"costUsd":53.56,"score":41.21,"level":"max"},"security":{"rank":24,"costUsd":27.66,"score":59.4,"level":"max","status":"partial"}},"levels":[{"level":"non-reasoning","codingScore":52.3,"pricePerMTokUsd":7,"thinkingMultiplier":0.3,"costUsd":30.88,"status":"below-quality-floor"},{"level":"low","codingScore":58.1,"pricePerMTokUsd":7,"thinkingMultiplier":0.43,"costUsd":24.87,"status":"below-quality-floor"},{"level":"medium","codingScore":64.7,"pricePerMTokUsd":7,"thinkingMultiplier":0.45,"costUsd":19.19,"status":"below-quality-floor"},{"level":"high","codingScore":67.1,"pricePerMTokUsd":7,"thinkingMultiplier":0.55,"costUsd":17.51,"status":"below-quality-floor"},{"level":"xhigh","codingScore":70.6,"pricePerMTokUsd":7,"thinkingMultiplier":1.16,"costUsd":15.76},{"level":"max","codingScore":76.7,"pricePerMTokUsd":7,"thinkingMultiplier":3.59,"costUsd":14.43}],"note":null},{"name":"Claude Opus 5","vendor":"Anthropic","releaseDate":"2026-07-24","basis":"measured","cursor":"Claude Opus 5","openrouter":"anthropic/claude-opus-5","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":32,"costUsd":14.82,"tokensUsd":5.29,"fixTimeUsd":9.52,"codingScore":76.5,"level":"high","pricePerMTokUsd":15},"byTask":{"greenfield":{"rank":16,"costUsd":16.96,"score":73.08,"level":"high"},"legacy":{"rank":6,"costUsd":26.21,"score":61.27,"level":"high","status":"partial"},"explore":{"rank":8,"costUsd":24.49,"score":63.17,"level":"high"},"debug":{"rank":8,"costUsd":14.94,"score":76.29,"level":"high"},"prod":{"rank":26,"costUsd":44.89,"score":46.19,"level":"high","status":"estimated"},"infra":{"rank":26,"costUsd":8.35,"score":89.07,"level":"high"},"algorithms":{"rank":31,"costUsd":29.64,"score":57.8,"level":"high"},"ui":{"rank":11,"costUsd":27.94,"score":59.47,"level":"high"},"security":{"rank":27,"costUsd":28.6,"score":58.81,"level":"high","status":"partial"}},"levels":[{"level":"low","codingScore":66.9,"pricePerMTokUsd":15,"thinkingMultiplier":0.35,"costUsd":19.82,"status":"below-quality-floor"},{"level":"medium","codingScore":74.3,"pricePerMTokUsd":15,"thinkingMultiplier":0.6,"costUsd":15.3},{"level":"high","codingScore":76.5,"pricePerMTokUsd":15,"thinkingMultiplier":1,"costUsd":14.82},{"level":"xhigh","codingScore":77,"pricePerMTokUsd":15,"thinkingMultiplier":1.7,"costUsd":15.99},{"level":"max","codingScore":78,"pricePerMTokUsd":15,"thinkingMultiplier":2.5,"costUsd":17.05}],"note":null},{"name":"Gemini 3.5 Flash","vendor":"Google","releaseDate":"2026-05-19","basis":"measured","cursor":"Gemini 3.5 Flash","openrouter":"google/gemini-3.5-flash","plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":33,"costUsd":15.24,"tokensUsd":2.02,"fixTimeUsd":13.22,"codingScore":70.1,"level":"high","pricePerMTokUsd":5.25},"byTask":{"greenfield":{"rank":41,"costUsd":49.44,"score":40.3,"level":"high","status":"partial"},"legacy":{"rank":37,"costUsd":90.19,"score":26.75,"level":"high","status":"partial"},"explore":{"rank":34,"costUsd":42.73,"score":43.97,"level":"high","status":"estimated"},"debug":{"rank":41,"costUsd":40.82,"score":45.14,"level":"high","status":"partial"},"prod":{"rank":37,"costUsd":49.44,"score":40.3,"level":"high"},"infra":{"rank":33,"costUsd":9.78,"score":79.49,"level":"high"},"algorithms":{"rank":30,"costUsd":29.36,"score":53.71,"level":"high","status":"partial"},"ui":{"rank":40,"costUsd":85.25,"score":27.89,"level":"high"},"security":{"rank":25,"costUsd":27.7,"score":55.23,"level":"high","status":"estimated"}},"levels":[{"level":"minimal","codingScore":null,"pricePerMTokUsd":null,"thinkingMultiplier":0.2,"costUsd":null,"status":"insufficient-benchmarks"},{"level":"medium","codingScore":null,"pricePerMTokUsd":null,"thinkingMultiplier":0.6,"costUsd":null,"status":"insufficient-benchmarks"},{"level":"high","codingScore":70.1,"pricePerMTokUsd":5.25,"thinkingMultiplier":1,"costUsd":15.24}],"note":null},{"name":"Claude Sonnet 5","vendor":"Anthropic","releaseDate":"2026-06-30","basis":"measured","cursor":"Claude Sonnet 5","openrouter":"anthropic/claude-sonnet-5","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":34,"costUsd":15.98,"tokensUsd":3.63,"fixTimeUsd":12.36,"codingScore":71.5,"level":"max","pricePerMTokUsd":6},"byTask":{"greenfield":{"rank":32,"costUsd":24.43,"score":60.61,"level":"max","status":"partial"},"legacy":{"rank":22,"costUsd":48.31,"score":42.36,"level":"max","status":"partial"},"explore":{"rank":40,"costUsd":66.77,"score":34.36,"level":"max","status":"estimated"},"debug":{"rank":37,"costUsd":30.18,"score":54.91,"level":"max","status":"partial"},"prod":{"rank":29,"costUsd":45.65,"score":43.83,"level":"max","status":"estimated"},"infra":{"rank":35,"costUsd":10.57,"score":80.81,"level":"max"},"algorithms":{"rank":36,"costUsd":30.98,"score":54.19,"level":"max"},"ui":{"rank":28,"costUsd":42.71,"score":45.57,"level":"max"},"security":{"rank":28,"costUsd":28.69,"score":56.27,"level":"max","status":"partial"}},"levels":[{"level":"non-reasoning · high","codingScore":66.4,"pricePerMTokUsd":6,"thinkingMultiplier":0.15,"costUsd":17.3,"status":"below-quality-floor"},{"level":"low","codingScore":58.4,"pricePerMTokUsd":6,"thinkingMultiplier":0.35,"costUsd":24.13,"status":"below-quality-floor"},{"level":"medium","codingScore":63.1,"pricePerMTokUsd":6,"thinkingMultiplier":0.6,"costUsd":20.28,"status":"below-quality-floor"},{"level":"high","codingScore":67.6,"pricePerMTokUsd":6,"thinkingMultiplier":1,"costUsd":17.25,"status":"below-quality-floor"},{"level":"xhigh","codingScore":69,"pricePerMTokUsd":6,"thinkingMultiplier":1.7,"costUsd":16.93,"status":"below-quality-floor"},{"level":"max","codingScore":71.5,"pricePerMTokUsd":6,"thinkingMultiplier":2.5,"costUsd":15.98}],"note":null},{"name":"Claude Fable 5.1","vendor":"Anthropic","releaseDate":"2026-09-01","basis":"measured","cursor":"Claude Fable 5.1","openrouter":"anthropic/claude-fable-5.1","plans":["claude-max","claude-team-premium","claude-enterprise","cursor","cursor-teams","cursor-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":35,"costUsd":16.75,"tokensUsd":8.56,"fixTimeUsd":8.19,"codingScore":79.1,"level":"high","pricePerMTokUsd":30},"byTask":{"greenfield":{"rank":24,"costUsd":20.15,"score":73.84,"level":"high"},"legacy":{"rank":9,"costUsd":34.05,"score":58.07,"level":"high"},"explore":{"rank":21,"costUsd":32.01,"score":59.95,"level":"high"},"debug":{"rank":16,"costUsd":17.93,"score":77.19,"level":"high"},"prod":{"rank":28,"costUsd":45.26,"score":49.53,"level":"high"},"infra":{"rank":36,"costUsd":10.65,"score":90.68,"level":"high"},"algorithms":{"rank":38,"costUsd":32.63,"score":59.36,"level":"high"},"ui":{"rank":12,"costUsd":28.2,"score":63.81,"level":"high"},"security":{"rank":37,"costUsd":37.39,"score":55.23,"level":"high"}},"levels":[{"level":"low","codingScore":75.2,"pricePerMTokUsd":30,"thinkingMultiplier":0.29,"costUsd":17.94},{"level":"medium","codingScore":77.1,"pricePerMTokUsd":30,"thinkingMultiplier":0.47,"costUsd":17.49},{"level":"high","codingScore":79.1,"pricePerMTokUsd":30,"thinkingMultiplier":0.59,"costUsd":16.75},{"level":"xhigh","codingScore":80.7,"pricePerMTokUsd":30,"thinkingMultiplier":2.34,"costUsd":22.83},{"level":"max","codingScore":81.6,"pricePerMTokUsd":30,"thinkingMultiplier":4.74,"costUsd":31.77}],"note":null},{"name":"GPT-5.4","vendor":"OpenAI","releaseDate":"2026-03-05","basis":"measured","cursor":null,"openrouter":"openai/gpt-5.4","plans":["copilot-pro","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":36,"costUsd":16.85,"tokensUsd":4.25,"fixTimeUsd":12.6,"codingScore":71.1,"level":"xhigh","pricePerMTokUsd":8.75},"byTask":{"greenfield":{"rank":39,"costUsd":31.04,"score":54.84,"level":"xhigh","status":"partial"},"legacy":{"rank":32,"costUsd":62.28,"score":36.47,"level":"xhigh","status":"partial"},"explore":{"rank":38,"costUsd":52.39,"score":40.8,"level":"xhigh"},"debug":{"rank":38,"costUsd":30.23,"score":55.56,"level":"xhigh","status":"partial"},"prod":{"rank":41,"costUsd":67.62,"score":34.5,"level":"xhigh"},"infra":{"rank":38,"costUsd":12.47,"score":78.28,"level":"xhigh","status":"partial"},"algorithms":{"rank":32,"costUsd":29.76,"score":56,"level":"xhigh","status":"partial"},"ui":{"rank":38,"costUsd":67.99,"score":34.37,"level":"xhigh"},"security":{"rank":31,"costUsd":30.29,"score":55.51,"level":"xhigh","status":"estimated"}},"levels":[{"level":"non-reasoning","codingScore":null,"pricePerMTokUsd":null,"thinkingMultiplier":0.15,"costUsd":null,"status":"insufficient-benchmarks"},{"level":"low","codingScore":null,"pricePerMTokUsd":null,"thinkingMultiplier":0.35,"costUsd":null,"status":"insufficient-benchmarks"},{"level":"xhigh","codingScore":71.1,"pricePerMTokUsd":8.75,"thinkingMultiplier":1.7,"costUsd":16.85}],"note":null},{"name":"GPT-6 Astra","vendor":"OpenAI","releaseDate":"2026-09-03","basis":"measured","cursor":null,"openrouter":"openai/gpt-6-astra","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":37,"costUsd":17.61,"tokensUsd":8.2,"fixTimeUsd":9.42,"codingScore":76.7,"level":"medium","pricePerMTokUsd":30},"byTask":{"greenfield":{"rank":22,"costUsd":18.22,"score":75.75,"level":"medium","status":"partial"},"legacy":{"rank":5,"costUsd":25.75,"score":65.7,"level":"medium"},"explore":{"rank":22,"costUsd":32.05,"score":59.14,"level":"medium"},"debug":{"rank":23,"costUsd":19.7,"score":73.55,"level":"medium"},"prod":{"rank":30,"costUsd":45.74,"score":48.59,"level":"medium"},"infra":{"rank":34,"costUsd":10.08,"score":89.66,"level":"low"},"algorithms":{"rank":37,"costUsd":31.64,"score":59.52,"level":"medium"},"ui":{"rank":5,"costUsd":25.42,"score":66.08,"level":"medium"},"security":{"rank":34,"costUsd":31.39,"score":59.76,"level":"medium"}},"levels":[{"level":"low","codingScore":75.7,"pricePerMTokUsd":30,"thinkingMultiplier":0.3,"costUsd":17.66},{"level":"medium","codingScore":76.7,"pricePerMTokUsd":30,"thinkingMultiplier":0.44,"costUsd":17.61},{"level":"high","codingScore":77.1,"pricePerMTokUsd":30,"thinkingMultiplier":1.07,"costUsd":20.01},{"level":"xhigh","codingScore":75.9,"pricePerMTokUsd":30,"thinkingMultiplier":1.93,"costUsd":24.49},{"level":"max","codingScore":76.9,"pricePerMTokUsd":30,"thinkingMultiplier":3.25,"costUsd":29.33}],"note":null},{"name":"GPT-5.5","vendor":"OpenAI","releaseDate":"2026-04-23","basis":"measured","cursor":null,"openrouter":"openai/gpt-5.5","plans":["chatgpt-plus","chatgpt-pro","chatgpt-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":38,"costUsd":17.91,"tokensUsd":5.55,"fixTimeUsd":12.36,"codingScore":71.5,"level":"medium","pricePerMTokUsd":17.5},"byTask":{"greenfield":{"rank":31,"costUsd":24.2,"score":63.35,"level":"medium","status":"partial"},"legacy":{"rank":28,"costUsd":51.28,"score":45.03,"level":"xhigh"},"explore":{"rank":37,"costUsd":44.87,"score":48.83,"level":"xhigh"},"debug":{"rank":31,"costUsd":26.4,"score":60.92,"level":"medium"},"prod":{"rank":25,"costUsd":44.29,"score":49.21,"level":"xhigh"},"infra":{"rank":37,"costUsd":12.02,"score":81.29,"level":"medium"},"algorithms":{"rank":35,"costUsd":30.78,"score":59.97,"level":"xhigh","status":"partial"},"ui":{"rank":36,"costUsd":57.54,"score":41.84,"level":"xhigh"},"security":{"rank":33,"costUsd":31.24,"score":59.53,"level":"xhigh","status":"partial"}},"levels":[{"level":"non-reasoning","codingScore":56.5,"pricePerMTokUsd":17.5,"thinkingMultiplier":0.15,"costUsd":29.39,"status":"below-quality-floor"},{"level":"low","codingScore":60.9,"pricePerMTokUsd":17.5,"thinkingMultiplier":0.35,"costUsd":25.64,"status":"below-quality-floor"},{"level":"medium","codingScore":71.5,"pricePerMTokUsd":17.5,"thinkingMultiplier":0.6,"costUsd":17.91},{"level":"high","codingScore":71.6,"pricePerMTokUsd":17.5,"thinkingMultiplier":1,"costUsd":18.9},{"level":"xhigh","codingScore":74.9,"pricePerMTokUsd":17.5,"thinkingMultiplier":1.7,"costUsd":18.46}],"note":null},{"name":"Claude Opus 4.8","vendor":"Anthropic","releaseDate":"2026-05-28","basis":"measured","cursor":"Claude Opus 4.8","openrouter":"anthropic/claude-opus-4.8","plans":["cursor","cursor-teams","cursor-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":39,"costUsd":19.44,"tokensUsd":8.72,"fixTimeUsd":10.72,"codingScore":74.3,"level":"max","pricePerMTokUsd":15},"byTask":{"greenfield":{"rank":36,"costUsd":27.64,"score":63.91,"level":"max","status":"partial"},"legacy":{"rank":25,"costUsd":50.13,"score":46.2,"level":"max"},"explore":{"rank":25,"costUsd":34.46,"score":57.26,"level":"max"},"debug":{"rank":36,"costUsd":29.32,"score":62.13,"level":"max"},"prod":{"rank":35,"costUsd":48.76,"score":46.99,"level":"max","status":"estimated"},"infra":{"rank":39,"costUsd":13.45,"score":84.33,"level":"max"},"algorithms":{"rank":39,"costUsd":37.55,"score":54.68,"level":"max","status":"partial"},"ui":{"rank":34,"costUsd":48.32,"score":47.25,"level":"max"},"security":{"rank":35,"costUsd":34.74,"score":57.01,"level":"max","status":"estimated"}},"levels":[{"level":"max","codingScore":74.3,"pricePerMTokUsd":15,"thinkingMultiplier":2.5,"costUsd":19.44}],"note":null},{"name":"Claude Opus 4.7","vendor":"Anthropic","releaseDate":"2026-04-16","basis":"measured","cursor":"Claude 4.7 Opus","openrouter":"anthropic/claude-opus-4.7","plans":["cursor","cursor-teams","cursor-enterprise"],"overall":{"rank":40,"costUsd":19.92,"tokensUsd":8.8,"fixTimeUsd":11.12,"codingScore":73.6,"level":"max","pricePerMTokUsd":15},"byTask":{"greenfield":{"rank":38,"costUsd":30.98,"score":60.48,"level":"max","status":"partial"},"legacy":{"rank":33,"costUsd":62.59,"score":40.05,"level":"max"},"explore":{"rank":39,"costUsd":61.96,"score":40.32,"level":"max"},"debug":{"rank":40,"costUsd":39.17,"score":53.41,"level":"max","status":"partial"},"prod":{"rank":36,"costUsd":49.33,"score":46.66,"level":"max"},"infra":{"rank":40,"costUsd":14.08,"score":83.15,"level":"max","status":"partial"},"algorithms":{"rank":40,"costUsd":38.66,"score":53.8,"level":"max","status":"partial"},"ui":{"rank":37,"costUsd":58.24,"score":42,"level":"max"},"security":{"rank":36,"costUsd":35.21,"score":56.61,"level":"max","status":"estimated"}},"levels":[{"level":"non-reasoning · high","codingScore":null,"pricePerMTokUsd":null,"thinkingMultiplier":0.15,"costUsd":null,"status":"insufficient-benchmarks"},{"level":"max","codingScore":73.6,"pricePerMTokUsd":15,"thinkingMultiplier":2.5,"costUsd":19.92}],"note":null},{"name":"Claude Fable 5","vendor":"Anthropic","releaseDate":"2026-06-09","basis":"measured","cursor":"Claude Fable 5","openrouter":"anthropic/claude-fable-5","plans":["claude-max","claude-team-premium","claude-enterprise","cursor","cursor-teams","cursor-enterprise","copilot-pro-plus","copilot-business","copilot-enterprise"],"overall":{"rank":41,"costUsd":26.46,"tokensUsd":16.94,"fixTimeUsd":9.52,"codingScore":76.5,"level":"max · opus 4.8 fallback","pricePerMTokUsd":30},"byTask":{"greenfield":{"rank":37,"costUsd":29.8,"score":72.31,"level":"max · opus 4.8 fallback","status":"partial"},"legacy":{"rank":18,"costUsd":43.42,"score":59.07,"level":"max · opus 4.8 fallback"},"explore":{"rank":41,"costUsd":81.72,"score":39,"level":"max · opus 4.8 fallback"},"debug":{"rank":39,"costUsd":31.09,"score":70.8,"level":"max · opus 4.8 fallback"},"prod":{"rank":39,"costUsd":64.18,"score":46.19,"level":"max · opus 4.8 fallback","status":"estimated"},"infra":{"rank":41,"costUsd":19.95,"score":86.27,"level":"max · opus 4.8 fallback"},"algorithms":{"rank":41,"costUsd":43.95,"score":58.65,"level":"max · opus 4.8 fallback","status":"partial"},"ui":{"rank":33,"costUsd":48.01,"score":55.64,"level":"max · opus 4.8 fallback"},"security":{"rank":38,"costUsd":44.44,"score":58.27,"level":"max · opus 4.8 fallback","status":"estimated"}},"levels":[{"level":"max · opus 4.8 fallback","codingScore":76.5,"pricePerMTokUsd":30,"thinkingMultiplier":2.5,"costUsd":26.46}],"note":null}],"disclaimer":"Data: Artificial Analysis and other public benchmarks, with their own licenses. Istari is independent and not affiliated with any model maker, Cursor (Anysphere), or Artificial Analysis."}