Skip to content

Benchmark

MathVision

Math benchmarks measure how well models solve maths problems. Scores are percentages. Higher is better.

Models measured
38
Scores recorded
47
Sources
2
IQ weight
1.95

Score distribution

76.1%
Mean
82.4%
Median
18%
Std dev
53.9%
10th percentile
94.5%
90th percentile
25%
Min
25%97.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.95 for chat models, measured over 37 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim97.8%
2LongCat 2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim96.5%
3Qwen 3.8 Flash Next· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.7%
4Qwen3.8 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.7%
5Qwen 3.8 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim94.6%
6Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim94.5%
7Kimi K2.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.2%
8Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim92.7%
9Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim90.3%
10Qwen3.7 PlusQwen3.7-Plus release · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.3%
11Qwen3.5 397B A17BQwen3.5 release/model card · fixed boxed-answer prompt· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.6%
12Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88%
13Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88%
14Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.2%
15Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86%
16Gemma 4 31B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.6%
17Gemma 4 31B ITGemma 4 Technical Report · Table 6 · thinking · 1120 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.6%
18Gemini 3.1 Pro PreviewSenseNova 6.7 Flash-Lite official benchmark chart · fixed boxed-answer prompt · peers use the higher reported boxed/unboxed run· Reported by SenseTime / OpenSenseNova· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.3%
19Kimi K2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.2%
20Kimi K2.5SenseNova 6.7 Flash-Lite official benchmark chart · fixed boxed-answer prompt · peers use the higher reported boxed/unboxed run· Reported by SenseTime / OpenSenseNova· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.2%
21Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.9%
22Gemma 4 31B ITGemma 4 Technical Report · Table 12 · thinking · 280 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.4%
23Gemma 4 26B A4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.4%
24Gemma 4 26B A4BGemma 4 Technical Report · Table 6 · thinking · 1120 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.4%
25GPT 5.4SenseNova 6.7 Flash-Lite official benchmark chart · fixed boxed-answer prompt · peers use the higher reported boxed/unboxed run· Reported by SenseTime / OpenSenseNova· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.4%
26Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.3%
27Gemma 4 26B A4BGemma 4 Technical Report · Table 12 · thinking · 280 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.3%
28Gemma 4 12B It· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.7%
29NVIDIA Nemotron 3.5 Super VL 120B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78.2%
30Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim78.1%
31Qwen3 VL 235Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.74.6%
32Qwen3.5 Omni PlusQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73%
33Claude Opus 4.6SenseNova 6.7 Flash-Lite official benchmark chart · fixed boxed-answer prompt · peers use the higher reported boxed/unboxed run· Reported by SenseTime / OpenSenseNova· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.72.1%
34Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.70.2%
35Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.66.5%
36qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.65.7%
37Qwen3.5 Omni FlashQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.65.4%
38Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.63.4%
39Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.62.7%
40qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.60.2%
41Gemma 4 E4B-it· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim59.5%
42Qwen3 VL 8B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.53.9%
43Gemma 4 E2B-it· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim52.4%
44Qwen2.5 VL 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim38.1%
45QVQ 72B Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim35.9%
46Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim25.1%
47Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim25%

Data credits: 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. · Data by LLM Stats