RepoQA: leaderboard
RepoQA: Evaluates software-engineering agents on realistic issue resolution, repository navigation, testing, or maintenance workflows.
Metric: Score (imported). Source: arxiv.org. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-05-13) | 90.6 | #325 |
| 2 | Gemini 1.5 Pro | 90.6 | #445 |
| 3 | Claude 3 Opus (20240229) | 90.6 | #412 |
| 4 | Gemini 1.5 Flash | 90 | #618 |
| 5 | Claude 3 Sonnet (20240229) | 87.4 | #677 |
| 6 | DeepSeek V2 Chat | 83.4 | #563 |
| 7 | Llama 3 70B Instruct | 82.2 | #634 |
| 8 | Claude 3 Haiku (20240307) | 81.8 | #771 |
| 9 | c4ai-command-r-plus | 78.4 | #724 |
| 10 | GPT-4 Turbo | 76.4 | #404 |
| 11 | Mixtral 8x7B Instruct (v0.1) | 68 | #1073 |
| 12 | Mixtral 8x22B Instruct (v0.1) | 67.8 | #756 |
| 13 | Qwen 1.5 72B Chat | 67 | #776 |
| 14 | Phi-3-medium-128k-instruct | 63.2 | #1013 |
| 15 | CodeQwen1.5-7B Chat | 62.8 | #1144 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=repoqa · How It Works · Data refreshed daily, snapshot 2026-10-20.