# FrontierCode 2.0

> Cognition's second-generation coding-agent benchmark, scoring whether agents produce mergeable pull requests on larger, less fully specified tasks, with added emphasis on thoroughness such as fixing every instance of a bug.

Canonical page: https://benchlm.ai/benchmarks/frontiercode-2

- Category: [Coding](/coding)
- Last updated: October 28, 2026 snapshot

## About FrontierCode 2.0

- Year: 2026
- Tasks: 132 private software-engineering tasks
- Format: Repository task completion with maintainer rubrics
- Difficulty: Frontier coding-agent quality
- Paper: [FrontierCode 2.0](https://cognition.com/blog/frontier-code-2.0)

FrontierCode 2.0 has 132 private tasks built by open-source maintainers from their own repositories. Compared with FrontierCode 1.1, tasks come from 55% more repositories, carry 43% more blocking criteria, and their reference solutions change 67% more files and 89% more lines. The score is a weighted aggregate of rubric items, and a solution that fails any blocking criterion scores zero. The mirrored table uses each model-agent row's best published reasoning effort; scores are not comparable with FrontierCode 1.1.

FrontierCode 2.0 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (20 models)

| Rank | Model | Configuration | Creator | Score |
|------|-------|---------------|---------|-------|
| 1 | [Claude Opus 5.5](/models/claude-opus-5-5) | claude-code · max reasoning | Anthropic | 45.5% |
| 2 | [GPT-6 Astra](/models/gpt-6-astra) | codex · max reasoning | OpenAI | 42.8% |
| 3 | [Claude Sonnet 5.5](/models/claude-sonnet-5-5) | claude-code · max reasoning | Anthropic | 41.0% |
| 4 | [GPT-6.1 Sol](/models/gpt-6-1-sol) | codex · max reasoning | OpenAI | 39.0% |
| 5 | [GPT-6 Sol](/models/gpt-6-sol) | codex · max reasoning | OpenAI | 34.3% |
| 6 | [Claude Fable 5.1](/models/claude-fable-5-1) | claude-code · xhigh reasoning | Anthropic | 32.4% |
| 7 | [Grok 4.7](/models/grok-4-7) | grok-build · xhigh reasoning | xAI | 28.6% |
| 8 | [Claude Haiku 5.5](/models/claude-haiku-5-5) | claude-code · max reasoning | Anthropic | 27.7% |
| 9 | [GPT-5.6 Sol](/models/gpt-5-6-sol) | codex · max reasoning | OpenAI | 27.6% |
| 10 | [Claude Opus 5](/models/claude-opus-5) | claude-code · xhigh reasoning | Anthropic | 26.7% |
| 11 | [Gemini 4 Argon](/models/gemini-4-argon) | devin-cli · high reasoning | Google | 26.3% |
| 12 | [Muse Wonder](/models/muse-wonder) | muse · high reasoning | Meta | 24.3% |
| 13 | [SWE-2](/models/swe-2) | devin · max reasoning | Cognition | 23.8% |
| 14 | [GPT-5.5](/models/gpt-5-5) | codex · xhigh reasoning | OpenAI | 22.6% |
| 15 | [GPT-6 Luna](/models/gpt-6-luna) | codex · max reasoning | OpenAI | 21.9% |
| 16 | [DeepSeek V4.1 Flash 1017](/models/deepseek-v4-1-flash) | devin-cli · max reasoning | DeepSeek | 19.3% |
| 17 | [Kimi K3](/models/kimi-k3) | devin-cli · high reasoning | Moonshot AI | 17.7% |
| 18 | [SWE-1.7](/models/swe-1-7) | devin-cli-swe1 | Cognition | 16.2% |
| 19 | [Claude Opus 4.8](/models/claude-opus-4-8) | claude-code · max reasoning | Anthropic | 14.7% |
| 20 | [SWE-1.6](https://cognition.com/blog/frontier-code-2.0) | devin-cli-swe1 | Cognition | 1.3% |

## FAQ

### What does FrontierCode 2.0 measure?

Cognition's second-generation coding-agent benchmark, scoring whether agents produce mergeable pull requests on larger, less fully specified tasks, with added emphasis on thoroughness such as fixing every instance of a bug.

### Which model leads the published FrontierCode 2.0 snapshot?

Claude Opus 5.5 currently leads the published FrontierCode 2.0 snapshot with a score of 45.5%.

### How many models are evaluated on FrontierCode 2.0?

The October 28, 2026 snapshot contains 20 AI models.

### Does FrontierCode 2.0 affect BenchLM's overall score?

Not directly. FrontierCode 2.0 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.