MMAnswerBench (BenchLM): leaderboard
Mathematical questions grounded in visual inputs such as diagrams and tables. BenchLM republishes model results attributed to Qwen’s launch evaluation. These are published results, not BenchLM ability estimates.
Metric: Accuracy (%). Source: benchlm.ai. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.2 | 91 | #80 |
| 2 | Kimi K2.6 | 86 | #113 |
| 3 | Claude Opus 4.5 | 84 | #87 |
| 4 | Qwen 3.6 Plus | 83.8 | #124 |
| 5 | GLM-5.1 | 83.8 | #109 |
| 6 | GLM-5 | 82.5 | #149 |
| 7 | Kimi K2.5 | 81.8 | #152 |
| 8 | Qwen 3.5 397B A17B | 80.9 | #154 |
| 9 | Qwen 3.6 27B | 80.8 | #182 |
| 10 | Qwen 3.6 35B A3B | 78.9 | #238 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmanswerbench-benchlm · How It Works · Data refreshed daily, snapshot 2026-11-05.