RepoMirage: leaderboard
Code agents are currently having skillful performance on repository-level software engineering benchmarks, but it remains unclear whether success on end-to-end tasks such as issue resolution truly reflects repository context reasoning, the ability to identify the task-relevant information across multiple files and reason over the relations among them. To investigate this question, we introduce Rep
Metric: Avg. (self-reported). Source: benchmarklist.com. Saturation forecast: Around August 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 41.4 | #58 |
| 2 | DeepSeek V3.2 | 39.8 | #205 |
| 3 | Claude Sonnet 4.6 | 28.2 | #90 |
| 4 | GPT-5 | 27.6 | #98 |
| 5 | Qwen 3.6 35B A3B | 24.2 | #226 |
| 6 | Qwen3 Coder Next | 22.6 | #326 |
| 7 | MiniMax-M2.7 | 14.8 | #258 |
| 8 | GPT-4.1 | 3.4 | #248 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=repomirage · How It Works · Data refreshed daily, snapshot 2026-10-20.