LLM2014 Code 2025-09: leaderboard
LLM2014 monthly coding benchmark for 2025-09, reporting the aggregate code-generation score across programming tasks.
Metric: Median Score. Source: raw.githubusercontent.com. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini (High) | 75.09 | #186 (GPT-5 Mini) |
| 2 | Claude Sonnet 4 (Thinking) | 61.01 | #199 (Claude Sonnet 4) |
| 3 | Claude Sonnet 4 | 51.35 | #199 |
| 4 | Gemini 2.5 Pro | 47.77 | #152 |
| 5 | DeepSeek V3.1 (Thinking) | 47.5 | #264 (DeepSeek V3.1) |
| 6 | Qwen 3 Coder 480B A35B | 38.89 | #279 |
| 7 | Grok 4 | 37.44 | #181 |
| 8 | GLM-4.5 (Thinking) | 37.21 | #275 (GLM-4.5) |
| 9 | Seed-1.6 (Thinking) | 36.7 | #265 (Seed-1.6) |
| 10 | Gemini 2.5 Flash (Thinking) | 36.34 | #245 (Gemini 2.5 Flash) |
| 11 | GPT-OSS-20B (High) | 36.2 | #518 (GPT-OSS-20B) |
| 12 | GLM-4.5 | 35.42 | #275 |
| 13 | Qwen 3 235B A22B 2507 Instruct | 33.22 | #300 |
| 14 | DeepSeek V3.1 | 32.69 | #264 |
| 15 | Kimi K2 (0711) | 27.7 | #492 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm2014-code-2025-09 · How It Works · Data refreshed daily, snapshot 2026-10-20.