Benchmark
MathVista
Math benchmarks measure how well models solve maths problems. Scores are percentages. Higher is better.
- Models measured
- 71
- Scores recorded
- 73
- Sources
- 3
- IQ weight
- 1.22
Score distribution
- 62.7%
- Mean
- 66.1%
- Median
- 14.8%
- Std dev
- 45%
- 10th percentile
- 79.2%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 1.22 for chat models, measured over 70 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 51 | Pixtral 12B2409· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 58% |
| 52 | Yi Vision· OpenVLM Leaderboard data | 57.8% |
| 53 | GPT-4o mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 56.7% |
| 54 | Pixtral 12B· OpenVLM Leaderboard data | 56.3% |
| 55 | GPT-4.1 nano· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 56.2% |
| 56 | Gemma 3 12B IT· OpenVLM Leaderboard data | 56.1% |
| 57 | Molmo 72B· OpenVLM Leaderboard data | 55.2% |
| 58 | Gemini 1.5 Flash-8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 54.7% |
| 59 | Llama 3.2 11B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 51.5% |
| 60 | Molmo 7B-D· OpenVLM Leaderboard data | 48.7% |
| 61 | Llama 3.2 11B Vision Instruct· OpenVLM Leaderboard data | 47.7% |
| 62 | gemini-1.0-pro· LLM Stats dataValues from 2 sources
| 46.6% |
| 63 | Gemma 3 4B· OpenVLM Leaderboard data | 46.3% |
| 64 | Claude 3 Opus· OpenVLM Leaderboard data | 46.1% |
| 65 | GLM 4 Vision· OpenVLM Leaderboard data | 45.9% |
| 66 | Claude 3 Sonnet· OpenVLM Leaderboard data | 45% |
| 67 | Phi-3.5 Vision Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 43.9% |
| 68 | Qwen-VL Max· OpenVLM Leaderboard data | 43.6% |
| 69 | Claude 3 Haiku· OpenVLM Leaderboard data | 42.6% |
| 70 | Qwen-VL Plus· OpenVLM Leaderboard data | 38.2% |
| 71 | Moondream2· OpenVLM Leaderboard data | 33.8% |
| 72 | kosmos-2· OpenVLM Leaderboard data | 24.5% |
| 73 | GPT-3.5 Turbo· LLM Stats data | 0% |
Showing 51–73 of 73 results.
Data credits: OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence) · Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.