FrontierCode 2.0 - Pass Rate
Code · 2026-10-27
FrontierCode 2.0 pass rate across Cognition's rebuilt suite of 132 production-code tasks. A trial passes only when it satisfies every blocker criterion and is not flagged for consulting solution-bearing sources. The result averages repeated trials per task. Scores are not comparable with FrontierCode 1.x.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 49.7 |
| GPT-6 Astra | 46.2 |
| Claude Sonnet 5.5 | 45.0 |
| GPT-6.1 Sol | 42.3 |
| GPT-6 Sol | 37.6 |
| Claude Fable 5.1 | 35.8 |
| Grok 4.7 | 32.0 |
| GPT-5.6 Sol | 30.6 |
| Claude Haiku 5.5 | 30.5 |
| Claude Opus 5 | 29.9 |
| Gemini 4 Argon | 29.1 |
| Muse Wonder | 27.5 |
| SWE-2 | 26.7 |
| GPT-5.5 | 25.3 |
| GPT-6 Luna | 24.2 |
| GLM-5.3 | 20.2 |
| Kimi K3 | 19.8 |
| Opus 4.8 | 16.4 |
| GLM-5.3 Flash | 15.9 |
| Inkling-Small | 3.8 |
| Beam 1.0 501B A23B | 3.3 |