IndustryCode - Python: leaderboard
Metric: Pass@1 (%) on the Python main problems of the IndustryCode test split (106 main problems decomposed into 499 sub-problems; industrial Python, C++, MATLAB and Stata projects with test cases), each main problem solved by implementing its sub-problems with cumulative context and integrating them; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 41.3 |
| 2 | Claude Sonnet 4.5 | 36.5 |
| 3 | Claude Opus 4.5 | 36.5 |
| 4 | Gemini 2.5 Pro | 34.9 |
| 5 | Qwen 3 Max | 31.7 |
| 6 | Qwen 3 Coder Plus | 31.7 |
| 7 | GLM-4.7 (Thinking) | 30.2 |
| 8 | Claude Haiku 4.5 | 28.6 |
| 9 | GLM-4.7 (Non-reasoning) | 28.6 |
| 10 | GPT-5.2 | 27 |
| 11 | GPT-5.1 | 27 |
| 12 | DeepSeek V3.2 Exp | 25.4 |
| 13 | Kimi K2 (Thinking) | 23.8 |
| 14 | Kimi K2 | 22.2 |
| 15 | Grok 4.1 | 22.2 |
Interactive version: theaggregate.ai/benchmark?slug=industrycode-python · How It Works · Data refreshed daily, snapshot 2026-10-07.