Skip to content

FrontierCode 1.1 Main

Cognition's 100-task software-engineering benchmark for whether coding agents produce mergeable, production-quality pull requests, scored for correctness, tests, scope, style, and maintainability through maintainer-authored rubrics.

16 models

Higher is better.

  1. 1Claude Opus 5.5Anthropic54.4
  2. 2Claude Fable 5Anthropic53.5
  3. 3Claude Opus 5Anthropic53.4
  4. 4GPT-6 AstraOpenAI53.3
  5. 5SWE-2Cognition50.0
  6. 6Claude Opus 4.8Anthropic46.5
  7. 7Claude Haiku 5.5Anthropic46.4
  8. 8Claude Sonnet 5.5Anthropic46.2
  9. 9Gemini 3.7 FlashGoogle43.6
  10. 10GPT-5.5OpenAI43.0
  11. 11Claude Sonnet 5Anthropic42.7
  12. 12SWE-1.7Cognition42.3
  13. 13Claude Opus 4.7Anthropic38.5
  14. 14GPT-5.4 miniOpenAI27.0
  15. 15Claude Opus 4.6Anthropic26.9
  16. 16Claude Sonnet 4.6Anthropic24.3
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Repository task completion with maintainer rubrics
Tasks
100 private Main tasks (150 in Extended)
Authors
Cognition (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.