Mathematics · Benchmark profile
A multimodal mathematical reasoning benchmark that tests whether models can answer visually grounded math questions correctly.
Data verified 21 Jul 2026 · Methodology 1.6.0
Benchmark score on MMAnswerBench
GLM-5.2 leads at 91%, followed by Kimi K2.6 (86%) and Claude Opus 4.5 (84%).
Visual analysis
Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.
10 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| GLM-5.2 | #1 | Z.AI | 91% |
| Kimi K2.6 | #2 | Moonshot AI | 86% |
| Claude Opus 4.5 | #3 | Anthropic | 84% |
| GLM-5.1 | #4 | Z.AI | 83.8% |
| Qwen3.6 Plus | #5 | Alibaba Cloud | 83.8% |
| GLM-5 | #6 | Z.AI | 82.5% |
| Kimi K2.5 | #7 | Moonshot AI | 81.8% |
| Qwen3.5 397B | #8 | Alibaba Cloud | 80.9% |
| Qwen3.6 27B | #9 | Alibaba Cloud | 80.8% |
| Qwen3.6-35B-A3B | #10 | Alibaba Cloud | 78.9% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | GLM-5.2 glm-5-2 | Z.AI | open | Estimated reference | 91% |
| #2 | Kimi K2.6 kimi-k2.6 | Moonshot AI | open | Estimated reference | 86% |
| #3 | Claude Opus 4.5 claude-opus-4-5 | Anthropic | closed | Reference only | 84% |
| #4 | GLM-5.1 glm-5-1 | Z.AI | open | Reference only | 83.8% |
| #5 | Qwen3.6 Plus qwen3-6-plus | Alibaba Cloud | closed | Reference only | 83.8% |
| #6 | GLM-5 glm-5 | Z.AI | open | Reference only | 82.5% |
| #7 | Kimi K2.5 kimi-k2.5 | Moonshot AI | — | Reference only | 81.8% |
| #8 | Qwen3.5 397B qwen3.5-397b-a17b | Alibaba Cloud | open | Estimated reference | 80.9% |
| #9 | Qwen3.6 27B qwen3.6-27b | Alibaba Cloud | open | Estimated reference | 80.8% |
| #10 | Qwen3.6-35B-A3B qwen3-6-35b-a3b | Alibaba Cloud | open | Estimated reference | 78.9% |
The top of this snapshot is led by GLM-5.2 at 91%; third place is 7.0 points behind. The top-10 spread is 12.1 points.
About MMAnswerBench
A multimodal mathematical reasoning benchmark that tests whether models can answer visually grounded math questions correctly. 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 multimodal mathematical reasoning benchmark that tests whether models can answer visually grounded math questions correctly.
GLM-5.2 by Z.AI currently leads with 91%.
10 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
