AA Coding Agent Index v2.0 - FrontierCode v1.1: leaderboard
FrontierCode v1.1 Main component of AA's Coding Agent Index v2.0: 100 of Cognition's 150 unreleased software-engineering tasks, solved by each model inside the coding-agent harness named in its row with internet access, three attempts per task; an attempt scores zero if a blocking check fails or it fetched the upstream solution, otherwise 70-100% rubric credit. Mean score (%).
Metric: Mean Score (%). Source: artificialanalysis.ai. Saturation forecast: Around 2030. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash (Max) | 52.6 | #37 (DeepSeek V4.1 Flash) |
| 2 | Gemini 4 Carbon (High) | 50.87 | #1 (Gemini 4 Carbon) |
| 3 | GPT-6.1 Sol (Max) | 49.41 | #7 (GPT-6.1 Sol) |
| 4 | GPT-6 Astra (Max) | 49.3 | #6 (GPT-6 Astra) |
| 5 | Muse Wonder Full (Max) | 48.73 | #12 (Muse Wonder Full) |
| 6 | Gemini 4 Argon (High) | 48.57 | #9 (Gemini 4 Argon) |
| 7 | GLM-5.3 (Max) | 48.25 | #44 (GLM-5.3) |
| 8 | Claude Haiku 5.5 (Max) | 47.37 | #43 (Claude Haiku 5.5) |
| 9 | Muse Spark 1.3 (Max) | 46.7 | #31 (Muse Spark 1.3) |
| 10 | DeepSeek V4.1 Flash (1017) (Max) | 46.55 | #27 (DeepSeek V4.1 Flash (1017)) |
| 11 | Muse Wonder (Max) | 46.39 | #18 (Muse Wonder) |
| 12 | GPT-6.1 Terra (xHigh) | 44.95 | #15 (GPT-6.1 Terra) |
| 13 | Qwen 3.8 Max | 44.81 | #34 |
| 14 | mistral-large-4-oss (High) | 44.55 | #60 (mistral-large-4-oss) |
| 15 | GPT-6.1 Terra (Medium) | 43.75 | #15 (GPT-6.1 Terra) |
No result here: #3 Claude Opus 5.5, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=aa-coding-agent-index-v2-0-frontiercode-v1-1 · How It Works · Data refreshed daily, snapshot 2026-11-05.