OpenVLM MathVista - ARI: leaderboard
OpenCompass OpenVLM MathVista arithmetic reasoning subcategory score.
Metric: Accuracy (%). Source: huggingface.co. Source: huggingface.co. Saturation forecast: Around January 2027. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 75.9 | #152 |
| 2 | InternVL3-78B | 74.5 | #352 |
| 3 | InternVL3-38B | 73.7 | #405 |
| 4 | GPT-5 | 72.2 | #98 |
| 5 | Doubao-1.5-Pro | 72 | #287 |
| 6 | GPT-5 Mini (2025-08-07) | 71.7 | #172 |
| 7 | InternVL2.5-78B-CoT | 68.8 | #469 |
| 8 | InternVL3-14B | 68 | #505 |
| 9 | Grok 2 (1212) | 67.1 | #496 |
| 10 | InternVL2.5-78B | 66.6 | #431 |
| 11 | GPT-4o ChatGPT | 66.6 | |
| 12 | GPT-4.1 (2025-04-14) | 65.7 | #250 |
| 13 | GPT-4.5 | 65.2 | #274 |
| 14 | InternVL2.5-26B-CoT | 64.9 | #642 |
| 15 | InternVL3-8B | 64.6 | #620 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mathvista-ari · How It Works · Data refreshed daily, snapshot 2026-10-20.