AdaptMMBench - Mode Selection MCC: leaderboard
Metric: Matthews correlation coefficient (-1 to 1) of adaptive-mode tool-invocation decisions against model-specific labels on 1,300 samples: a sample is tool-required when the model's own text-mode answer is wrong and tool-redundant when it is right (0 is chance; a model that always or never calls tools scores about 0); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 0.41 | #91 |
| 2 | Qwen 3 VL 235B A22B Instruct | 0.26 | #264 |
| 3 | Gemini 3 Pro | 0.24 | #77 |
| 4 | Qwen 3 VL 32B Instruct | 0.14 | #276 |
| 5 | Qwen 3 VL 8B Instruct | 0.06 | #401 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=adaptmmbench-mode-selection-mcc · How It Works · Data refreshed daily, snapshot 2026-10-11.