Multi-SWE-bench flash (MopenHands): leaderboard
Multi-SWE-bench flash, a subset of 300 issues that the Multi-SWE-bench maintainers (ByteDance Seed, 2025) selected for fast evaluation; the agent resolves real GitHub issues by editing the repository, and a fix counts only when the project's tests pass. Scored as the percentage of issues resolved under MopenHands, the benchmark team's multilingual port of OpenHands; higher is better.
Metric: Resolved rate (%). Source: multi-swe-bench.github.io. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Coder 480B A35B | 25.33 | #677 |
| 2 | Claude Sonnet 4 | 25 | #397 |
| 3 | O3 | 23.67 | #224 |
| 4 | Claude 3.7 Sonnet (20250219) | 21.33 | #399 |
| 5 | Kimi K2 | 18.33 | #435 |
| 6 | Gemini 2.5 Pro | 15.67 | #262 |
| 7 | DeepSeek R1 0528 | 15.33 | #425 |
| 8 | GPT-4.1 | 14.33 | #492 |
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-flash-mopenhands · How It Works · Data refreshed daily, snapshot 2026-11-12.