MMErroR - Error Type Classification - Physics & Engineering: leaderboard
Metric: Error Type Classification Accuracy (%). Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Pro | 67.32 |
| 2 | Gemini 3 Pro (Preview) | 66.88 |
| 3 | GPT-5.2 (xHigh) | 63.62 |
| 4 | Claude Opus 4.5 | 61 |
| 5 | Grok 4 | 55.56 |
| 6 | Kimi K2.5 | 55.56 |
| 7 | Qwen 3 VL 32B (Thinking) | 54.9 |
| 8 | GLM-4.6V | 41.18 |
| 9 | Llama 4 Maverick Instruct | 36.17 |
| 10 | Llama 4 Scout Instruct | 34.86 |
| 11 | Qwen 3 VL 32B Instruct | 32.68 |
Interactive version: theaggregate.ai/benchmark?slug=mmerror-error-type-classification-physics-engineering · How It Works · Data refreshed daily, snapshot 2026-09-25.