Multi-SWE-bench (MopenHands): leaderboard
Multi-SWE-bench (ByteDance Seed, 2025) asks an agent to resolve real GitHub issues by editing the repository; a fix counts only when the project's own tests pass. This board is the official leaderboard's All view under MopenHands, the benchmark team's multilingual port of OpenHands: issues resolved across all eight language splits (Java, TypeScript, JavaScript, Go, Rust, C, C++ and the 500 Python tasks of SWE-bench Verified; 2,132 issues), in percent, for models that ran every split. Higher is better.
Metric: Resolved rate (%). Source: multi-swe-bench.github.io. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 21.62 | #234 |
| 2 | Claude 3.7 Sonnet (20250219) | 19.32 | #399 |
| 3 | Claude 3.5 Sonnet (20241022) | 15.24 | #635 |
| 4 | DeepSeek V3 | 8.72 | #705 |
| 5 | GPT-4o (2024-11-20) | 8.21 | #864 |
| 6 | DeepSeek R1 | 8.02 | #467 |
| 7 | O3 Mini (2025-01-31) (High) | 7.55 | #486 (O3 Mini (2025-01-31)) |
| 8 | Llama 4 Maverick | 7.46 | #1105 |
| 9 | O1 (2024-12-17) | 6.1 | #310 |
| 10 | Doubao-1.5-Pro | 2.91 | #585 |
| 11 | Qwen 2.5 72B Instruct | 2.02 | #1093 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=multi-swe-bench-mopenhands · How It Works · Data refreshed daily, snapshot 2026-11-12.