ProLLM - StackUnseen: leaderboard
ProLLM benchmark evaluating LLMs on unseen Stack Overflow coding questions not in training data. Measures acceptance rate on real developer problems.
Metric: Score (%). Source: www.prollm.ai. Saturation forecast: Around January 2027. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 88.6 | #110 |
| 2 | Grok 4 | 88.6 | #181 |
| 3 | Gemini 3 Pro (Preview) | 86.2 | #66 |
| 4 | Gemini 3 Flash | 83 | #99 |
| 5 | GPT-5 Mini | 82.4 | #186 |
| 6 | Claude Opus 4.5 | 81.9 | #83 |
| 7 | Qwen 3.5 397B A17B | 76.3 | #146 |
| 8 | Kimi K2 (Thinking) | 76.1 | #243 (Kimi K2) |
| 9 | MiniMax-M2.5 | 66 | #303 |
| 10 | Kimi K2.5 | 64.9 | #144 |
| 11 | GPT-5 Nano | 60.4 | #425 |
| 12 | GLM-5 | 55.1 | #141 |
| 13 | DeepSeek R1 0528 | 52.4 | #222 |
| 14 | Mistral Large 3 | 51.6 | #401 |
| 15 | DeepSeek V3.1 | 48.1 | #264 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=prollm-stackunseen · How It Works · Data refreshed daily, snapshot 2026-10-20.