GR-Ben (Thinking) - Computer Science: leaderboard
Metric: F1 (%) of first-error identification: the harmonic mean of the accuracy on flawed solutions (the earliest erroneous step named) and the accuracy on fully correct solutions (judged correct), on the 400 computer-science problems (MMLU-Pro) of GR-Ben (3,595 human-annotated solutions written by 15 LLMs); LLMs prompted as critics with slow thinking enabled (GPT-5.2 at xhigh and Gemini 3 Flash at high reasoning effort, 32,768 output tokens); higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (High) | 66.5 |
| 2 | DeepSeek V3.2 (Thinking) | 59.5 |
| 3 | GPT-5.2 (xHigh) | 56.3 |
| 4 | Qwen 3 32B | 47.6 |
| 5 | Qwen 3 14B | 47.3 |
| 6 | Qwen 3 8B | 38.2 |
| 7 | Qwen 3 4B | 34.8 |
Interactive version: theaggregate.ai/benchmark?slug=gr-ben-thinking-computer-science · How It Works · Data refreshed daily, snapshot 2026-10-07.