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.
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)).
| Model | Lab | Score↓ | Source model | Status | Date | |
|---|---|---|---|---|---|---|
| Claude Fable 5.2 | Anthropic | 50.2% | Claude Fable 5.2 | Official | Nov 9, 2026 | Details |
| Claude Opus 5.5 | Anthropic | 47.3% | Claude Opus 5.5 | Official | Sep 22, 2026 | Details |
| GPT-6 Astra | OpenAI | 44.0% | GPT-6 Astra | Official | Sep 3, 2026 | Details |
| Gemini 4 Carbon | Google DeepMind | 42.6% | Gemini 4 Carbon | Official | Nov 10, 2026 | Details |
| Claude Sonnet 5.5 | Anthropic | 42.2% | Claude Sonnet 5.5 | Official | Sep 28, 2026 | Details |
| GPT-6.1 Sol | OpenAI | 39.6% | GPT-6.1 Sol | Official | Sep 29, 2026 | Details |
| Gemini 4 Hydrogen | Google DeepMind | 38.9% | Gemini 4 Hydrogen | Official | Nov 12, 2026 | Details |
| GPT-6 Sol | OpenAI | 34.9% | GPT-6 Sol | Official | Sep 22, 2026 | Details |
| Claude Fable 5.1 | Anthropic | 34.0% | Claude Fable 5.1 | Official | Sep 1, 2026 | Details |
| Claude Haiku 5.5 | Anthropic | 28.9% | Claude Haiku 5.5 | Official | Oct 7, 2026 | Details |
| Grok 4.7 | xAI | 28.9% | Grok 4.7 | Official | Sep 21, 2026 | Details |
| GPT-5.6 Sol | OpenAI | 28.9% | GPT-5.6 Sol | Official | Jul 9, 2026 | Details |
| Claude Opus 5 | Anthropic | 28.1% | Claude Opus 5 | Official | Jul 24, 2026 | Details |
| Gemini 4 Argon | Google DeepMind | 27.6% | Gemini 4 Argon | Official | Oct 13, 2026 | Details |
| Muse Wonder | Meta | 25.3% | Muse Wonder 1.0 | Official | Oct 12, 2026 | Details |
| SWE-2 | Cognition | 24.1% | SWE-2 | Official | Sep 10, 2026 | Details |
| SWE-2 Flash | Cognition | 24.0% | SWE-2 Flash | Official | Nov 4, 2026 | Details |
| GPT-5.5 | OpenAI | 23.6% | GPT-5.5 | Official | Apr 23, 2026 | Details |
| GPT-6 Luna | OpenAI | 22.7% | GPT-6 Luna | Official | Sep 22, 2026 | Details |
| Inkling 2 SmallOpen | Thinking Machines Lab | 22.6% | Inkling 2 Small | Official | Nov 5, 2026 | Details |
| DeepSeek V4.1 Flash 1017Open | DeepSeek | 20.2% | DeepSeek V4.1 Flash 1017 | Official | Oct 17, 2026 | Details |
| GLM-5.3Open | Z.ai | 20.2% | GLM 5.3 | Official | Aug 14, 2026 | Details |
| Kimi K3.1 ProOpen | Moonshot AI | 19.9% | Kimi K3.1 Pro | Official | Oct 23, 2026 | Details |
| Mistral Large 4Open | Mistral AI | 19.7% | Mistral Large 4 | Official | Oct 29, 2026 | Details |
| Kimi K3Open | Moonshot AI | 18.9% | Kimi K3 | Official | Jul 16, 2026 | Details |
| SWE-1.7 | Cognition | 15.5% | SWE-1.7 | Official | Jul 8, 2026 | Details |
| Claude Opus 4.8 | Anthropic | 15.3% | Claude Opus 4.8 | Official | May 28, 2026 | Details |
| GLM-5.3-FlashOpen | Z.ai | 14.9% | GLM 5.3 Flash | Official | Aug 26, 2026 | Details |
| Muse Spark OpenOpen | Meta | 6.5% | Muse Spark Open | Official | Oct 15, 2026 | Details |
| Inkling SmallOpen | Thinking Machines Lab | 4.0% | Inkling Small | Official | Jul 30, 2026 | Details |
| Beam 1.0Open | Reflection AI | 3.4% | Beam 1.0 | Official | Oct 16, 2026 | Details |
| SWE-1.6 | Cognition | 1.3% | SWE-1.6 | Official | Apr 7, 2026 | Details |
Each row reports the model’s weighted score (Main) on FrontierCode 2.0. Click a row for the full run context.