math benchmark
PolyMATH Leaderboard
PolyMATH is a multimodal mathematical reasoning benchmark for evaluating Multi-modal Large Language Models (MLLMs) on textual and visual cognitive challenges.
Updated Sep 5, 2026
Models23
Model coverage23
MetricScore
EvidenceB
PolyMATH Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score86.50% | Percentile100.00% | Participants23 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score84.00% | Percentile95.45% | Participants23 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score77.40% | Percentile90.91% | Participants23 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score73.30% | Percentile86.36% | Participants23 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score71.20% | Percentile81.82% | Participants23 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score68.90% | Percentile77.27% | Participants23 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score64.40% | Percentile72.73% | Participants23 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score60.10% | Percentile68.18% | Participants23 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score57.30% | Percentile63.64% | Participants23 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score56.30% | Percentile59.09% | Participants23 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score52.00% | Percentile54.55% | Participants23 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score51.70% | Percentile50.00% | Participants23 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score51.10% | Percentile45.45% | Participants23 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score50.20% | Percentile40.91% | Participants23 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score47.50% | Percentile36.36% | Participants23 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score45.90% | Percentile31.82% | Participants23 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score44.60% | Percentile27.27% | Participants23 | EvidenceC | Evaluated |
| Rank18 | ModelAC | Score44.30% | Percentile22.73% | Participants23 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score40.50% | Percentile18.18% | Participants23 | EvidenceC | Evaluated |
| Rank20 | ModelAC | Score30.40% | Percentile13.64% | Participants23 | EvidenceC | Evaluated |
| Rank21 | ModelAC | Score28.80% | Percentile9.09% | Participants23 | EvidenceC | Evaluated |
| Rank22 | ModelAC | Score26.10% | Percentile4.55% | Participants23 | EvidenceC | Evaluated |
| Rank23 | ModelAC | Score8.20% | Percentile0.00% | Participants23 | EvidenceC | Evaluated |
PolyMATH Highlights
The leading models and scores on this benchmark.
PolyMATH Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for PolyMATH
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is PolyMATH?
What PolyMATH measures and how its scores work.
PolyMATH comprises 5,000 manually collected images across 10 categories, including pattern recognition, spatial reasoning, and relative reasoning.
It measures performance on multimodal mathematical and cognitive reasoning challenges, reported as Score in ratio units.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- PolyMATH
- Modality
- multimodal
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about PolyMATH.
Which model scores highest on PolyMATH?
Qwen3.7 Max is currently ranked first with 86.50%.
What are the top three models on PolyMATH?
The current leaders are Qwen3.7 Max (86.50%), Qwen3.7-Plus (84.00%), and Qwen3.6 Plus (77.40%).
Which PolyMATH model has the lowest official input price?
Qwen3.5-35B-A3B has the lowest matched official input price at $0.25 input / $2.0 output per 1M tokens.
Which models are fastest among PolyMATH results?
The fastest matched records are Qwen3 VL 4B Thinking (8.77 tok/s via DeepInfra).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.