GitTaskBench (OpenHands): leaderboard
GitTaskBench (Ni et al., 2025) gives a code agent a real-world task, such as restoring an old photo or extracting data from a PDF, plus the GitHub repository it must set up and use to do it: 54 tasks from 18 repositories across 7 domains. This board is the task pass rate under OpenHands: the percentage of tasks whose output meets the task's success criteria, averaged over two runs; higher is better.
Metric: Task pass rate (%). Source: gittaskbench.github.io. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 48.15 | #399 |
| 2 | GPT-4.1 | 42.59 | #492 |
| 3 | Claude 3.5 Sonnet (20241022) | 40.74 | #635 |
| 4 | Gemini 2.5 Pro | 35.19 | #262 |
| 5 | Qwen 3 32B (Thinking) | 29.63 | #960 (Qwen 3 32B) |
| 6 | DeepSeek V3 (0324) | 26.85 | #704 |
| 7 | Qwen 3 32B (Non-reasoning) | 25.93 | #960 (Qwen 3 32B) |
| 8 | O3 Mini (Medium) | 22.22 | #536 (O3 Mini) |
| 9 | Llama 3.3 70B Instruct | 20.37 | #1353 |
| 10 | GPT-4o (2024-08-06) | 14.82 | #731 |
| 11 | Qwen 3 14B (Non-reasoning) | 5.56 | #1235 (Qwen 3 14B) |
| 12 | Qwen 3 8B (Non-reasoning) | 1.85 | #1838 (Qwen 3 8B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=gittaskbench-openhands · How It Works · Data refreshed daily, snapshot 2026-11-12.