Multimodal · Benchmark profile
A visual mathematics benchmark that tests whether a model can solve math problems grounded in diagrams, equations, figures, and other visual inputs.
Data verified 21 Jul 2026 · Methodology 1.6.0
Benchmark score on MathVision
Kimi K3 leads at 94.3%, followed by Qwen3.7-Plus (90.3%) and Qwen3.5 397B (88.6%).
Visual analysis
Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.
12 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Kimi K3 | #1 | Moonshot AI | 94.3% |
| Qwen3.7-Plus | #2 | Alibaba Cloud | 90.3% |
| Qwen3.5 397B | #3 | Alibaba Cloud | 88.6% |
| Qwen3.6 Plus | #4 | Alibaba Cloud | 88% |
| Kimi K2.6 | #5 | Moonshot AI | 87.4% |
| Gemini 3 Pro | #6 | 86.6% | |
| Qwen3.5-122B-A10B | #7 | Alibaba Cloud | 86.2% |
| Qwen3.5 27B | #8 | Alibaba Cloud | 86% |
| Qwen3.5-35B-A3B | #9 | Alibaba Cloud | 83.9% |
| GPT-5.2 | #10 | OpenAI | 83% |
| Gemma 4 12B | #11 | 79.7% | |
| Claude Opus 4.5 | #12 | Anthropic | 74.3% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | Kimi K3 kimi-k3 | Moonshot AI | closed | Reference only | 94.3% |
| #2 | Qwen3.7-Plus qwen3.7-plus | Alibaba Cloud | — | Reference only | 90.3% |
| #3 | Qwen3.5 397B qwen3.5-397b-a17b | Alibaba Cloud | open | Estimated reference | 88.6% |
| #4 | Qwen3.6 Plus qwen3-6-plus | Alibaba Cloud | closed | Reference only | 88% |
| #5 | Kimi K2.6 kimi-k2.6 | Moonshot AI | open | Estimated reference | 87.4% |
| #6 | Gemini 3 Pro gemini-3-pro | closed | Reference only | 86.6% | |
| #7 | Qwen3.5-122B-A10B qwen3-5-122b-a10b | Alibaba Cloud | open | Reference only | 86.2% |
| #8 | Qwen3.5 27B qwen3.5-27b | Alibaba Cloud | open | Reference only | 86% |
| #9 | Qwen3.5-35B-A3B qwen3-5-35b-a3b | Alibaba Cloud | open | Reference only | 83.9% |
| #10 | GPT-5.2 gpt-5.2 | OpenAI | closed | Estimated reference | 83% |
| #11 | Gemma 4 12B gemma-4-12b | open | Estimated reference | 79.7% | |
| #12 | Claude Opus 4.5 claude-opus-4-5 | Anthropic | closed | Reference only | 74.3% |
The top of this snapshot is led by Kimi K3 at 94.3%; third place is 5.7 points behind. The top-10 spread is 11.3 points.
About MathVision
A visual mathematics benchmark that tests whether a model can solve math problems grounded in diagrams, equations, figures, and other visual inputs. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
A visual mathematics benchmark that tests whether a model can solve math problems grounded in diagrams, equations, figures, and other visual inputs.
Kimi K3 by Moonshot AI currently leads with 94.3%.
12 models in the LuminaBench cohort have a qualifying score on this benchmark.
No. This benchmark is display-only and does not enter the overall Lumina composite.
Related
