GosuEvals: leaderboard
Coding evaluations of explicit model–agent combinations across multi-file projects. Scores combine judge ratings, unit tests and deterministic checks; source test dates and thinking settings are retained.
Metric: Score (%). Source: gosuevals.com. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4 (Thinking) | 74.3 | #199 (Claude Sonnet 4) |
| 2 | Claude Sonnet 4 | 71.8 | #199 |
| 3 | Claude 3.7 Sonnet | 71.8 | #249 |
| 4 | Claude Opus 4 | 71.8 | #163 |
| 5 | Claude 3.7 Sonnet (Thinking) | 71 | #249 (Claude 3.7 Sonnet) |
| 6 | Claude 3.5 Sonnet | 69.5 | #350 |
| 7 | Gemini 2.5 Pro (Preview 06-05) | 68.5 | #106 |
| 8 | DeepSeek R1 0528 | 64.8 | #222 |
| 9 | Gemini 2.5 Pro (Preview 03-25) | 63.1 | #123 |
| 10 | Gemini 2.5 Pro (Preview) | 58.3 | #130 |
| 11 | Gemini 2.5 Flash (Preview 05-20) | 56.2 | #216 |
| 12 | Gemini 2.5 Pro (Preview 05-06) | 52 | #147 |
| 13 | Gemini 2.0 Flash | 49 | #342 |
| 14 | Grok 3 Mini (High) | 47.5 | #281 (Grok 3 Mini) |
| 15 | GPT-4.1 | 41 | #248 |
Interactive version: theaggregate.ai/benchmark?slug=gosuevals · How It Works · Data refreshed daily, snapshot 2026-10-20.