GitTaskBench (Aider): 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 the Aider coding assistant: 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. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 (0324) | 16.67 | #704 |
| 2 | Claude 3.5 Sonnet (20241022) | 12.96 | #635 |
| 3 | GPT-4.1 | 7.41 | #492 |
| 4 | GPT-4o (2024-08-06) | 1.85 | #731 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=gittaskbench-aider · How It Works · Data refreshed daily, snapshot 2026-11-12.