MicroVQA - Expert Visual Understanding: leaderboard
Accuracy (%) on the MicroVQA questions of the expert visual understanding task: interpreting what a research microscopy image shows. Each model answers every question of this task once with the same chain-of-thought prompt as the headline board.
Metric: Accuracy (%). Source: jmhb0.github.io. Saturation forecast: Estimated already saturated. 14 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 (2024-12-17) | 55.4 | #314 |
| 2 | Qwen 2 VL 7B Instruct | 54.1 | #2594 |
| 3 | pixtral-large-2411 | 50.8 | #1284 |
| 4 | Grok 2 (1212) | 50.3 | #1082 |
| 5 | Qwen 2 VL 72B Instruct | 49.2 | #1382 |
| 6 | Gemini 1.5 Flash-8B (001) | 48.7 | #2311 |
| 7 | GPT-4o Mini (2024-07-18) | 48.5 | #1494 |
| 8 | Claude 3.5 Haiku (20241022) | 48 | #1589 |
| 9 | VILA1.5-40B | 47.2 | #3238 |
| 10 | Pixtral-12B-2409 | 46.9 | #2537 |
| 11 | Llama 3.2 90B Vision Instruct | 44.9 | #1791 |
| 12 | Llama 3.1 Nemotron 70B Instruct | 44.9 | #1471 |
| 13 | VILA1.5-13B | 41.8 | #7029 |
| 14 | Llama 3.2 11B Instruct | 32.4 | #5226 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=microvqa-expert-visual-understanding · How It Works · Data refreshed daily, snapshot 2026-11-17.