MA-Bench - Coarse-Grained Recognition: leaderboard
Metric: Accuracy (%) on the coarse-grained (body-level) micro-action recognition questions (random 14.7), MA-Bench's 1,000 micro-action videos (about 2 seconds each, 52 micro-action categories), 8 sampled frames, zero-shot, final answer without intermediate reasoning; higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 43 | #237 |
| 2 | Qwen 2 VL 7B | 28.5 | #741 |
| 3 | GPT-4o | 20.5 | #333 |
| 4 | Qwen 2.5 VL 7B Instruct | 20.4 | #643 |
| 5 | InternVL2-8B | 17.4 | #826 |
| 6 | InternVL3-8B | 17.1 | #606 |
| 7 | Pixtral-12B | 14.4 | #795 |
| 8 | Phi-4 Multimodal Instruct | 11.4 | #896 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=ma-bench-coarse-grained-recognition · How It Works · Data refreshed daily, snapshot 2026-10-11.