evals.report
BenchmarksLabsCompareRun guidesIn the wild

FrontierCode 2.0

Cognition's second-generation code-mergeability benchmark (released 2026-10-27): 132 private tasks crafted by the open-source maintainers of the repositories they come from, graded against maintainer-defined rubrics where any failed blocker criterion scores the solution 0. Compared with FrontierCode 1.1, tasks come from 55% more repositories and add categories such as security patches and performance optimizations; reference solutions touch 67% more files and 89% more lines in codebases 25% larger; and tasks carry 43% more blocking criteria, with a new emphasis on 'thoroughness' (fixing every instance of a bug, matching existing support for related inputs). In 80% of tasks the reference solution was corrected from the upstream PR. Each model is run 5× at every available reasoning effort; runs that consulted solution-bearing sources are zeroed.

Codingweighted score (Main)Higher is better

What is FrontierCode 2.0?

Cognition's second-generation code-mergeability benchmark (released 2026-10-27): 132 private tasks crafted by the open-source maintainers of the repositories they come from, graded against maintainer-defined rubrics where any failed blocker criterion scores the solution 0. Compared with FrontierCode 1.1, tasks come from 55% more repositories and add categories such as security patches and performance optimizations; reference solutions touch 67% more files and 89% more lines in codebases 25% larger; and tasks carry 43% more blocking criteria, with a new emphasis on 'thoroughness' (fixing every instance of a bug, matching existing support for related inputs). In 80% of tasks the reference solution was corrected from the upstream PR. Each model is run 5× at every available reasoning effort; runs that consulted solution-bearing sources are zeroed. evals.report tracks reported FrontierCode 2.0 scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported FrontierCode 2.0 score: Claude Fable 5.2 — 50.2% (weighted score (Main)).

ModelLabScore↓Source modelStatusDate
Claude Fable 5.2Anthropic50.2%Claude Fable 5.2OfficialNov 9, 2026Details
Claude Opus 5.5Anthropic47.3%Claude Opus 5.5OfficialSep 22, 2026Details
GPT-6 AstraOpenAI44.0%GPT-6 AstraOfficialSep 3, 2026Details
Gemini 4 CarbonGoogle DeepMind42.6%Gemini 4 CarbonOfficialNov 10, 2026Details
Claude Sonnet 5.5Anthropic42.2%Claude Sonnet 5.5OfficialSep 28, 2026Details
GPT-6.1 SolOpenAI39.6%GPT-6.1 SolOfficialSep 29, 2026Details
Gemini 4 HydrogenGoogle DeepMind38.9%Gemini 4 HydrogenOfficialNov 12, 2026Details
GPT-6 SolOpenAI34.9%GPT-6 SolOfficialSep 22, 2026Details
Claude Fable 5.1Anthropic34.0%Claude Fable 5.1OfficialSep 1, 2026Details
Claude Haiku 5.5Anthropic28.9%Claude Haiku 5.5OfficialOct 7, 2026Details
Grok 4.7xAI28.9%Grok 4.7OfficialSep 21, 2026Details
GPT-5.6 SolOpenAI28.9%GPT-5.6 SolOfficialJul 9, 2026Details
Claude Opus 5Anthropic28.1%Claude Opus 5OfficialJul 24, 2026Details
Gemini 4 ArgonGoogle DeepMind27.6%Gemini 4 ArgonOfficialOct 13, 2026Details
Muse WonderMeta25.3%Muse Wonder 1.0OfficialOct 12, 2026Details
SWE-2Cognition24.1%SWE-2OfficialSep 10, 2026Details
SWE-2 FlashCognition24.0%SWE-2 FlashOfficialNov 4, 2026Details
GPT-5.5OpenAI23.6%GPT-5.5OfficialApr 23, 2026Details
GPT-6 LunaOpenAI22.7%GPT-6 LunaOfficialSep 22, 2026Details
Inkling 2 SmallOpenThinking Machines Lab22.6%Inkling 2 SmallOfficialNov 5, 2026Details
DeepSeek V4.1 Flash 1017OpenDeepSeek20.2%DeepSeek V4.1 Flash 1017OfficialOct 17, 2026Details
GLM-5.3OpenZ.ai20.2%GLM 5.3OfficialAug 14, 2026Details
Kimi K3.1 ProOpenMoonshot AI19.9%Kimi K3.1 ProOfficialOct 23, 2026Details
Mistral Large 4OpenMistral AI19.7%Mistral Large 4OfficialOct 29, 2026Details
Kimi K3OpenMoonshot AI18.9%Kimi K3OfficialJul 16, 2026Details
SWE-1.7Cognition15.5%SWE-1.7OfficialJul 8, 2026Details
Claude Opus 4.8Anthropic15.3%Claude Opus 4.8OfficialMay 28, 2026Details
GLM-5.3-FlashOpenZ.ai14.9%GLM 5.3 FlashOfficialAug 26, 2026Details
Muse Spark OpenOpenMeta6.5%Muse Spark OpenOfficialOct 15, 2026Details
Inkling SmallOpenThinking Machines Lab4.0%Inkling SmallOfficialJul 30, 2026Details
Beam 1.0OpenReflection AI3.4%Beam 1.0OfficialOct 16, 2026Details
SWE-1.6Cognition1.3%SWE-1.6OfficialApr 7, 2026Details

Each row reports the model’s weighted score (Main) on FrontierCode 2.0. Click a row for the full run context.