Benchmark
PolyMATH Benchmark
What is PolyMATH?
Polymath is a challenging multi-modal mathematical reasoning benchmark designed to evaluate the general cognitive reasoning abilities of Multi-modal Large…
- Modality
- multimodal
- Evaluates
- math, multimodal, reasoning, spatial reasoning, vision
- Openness
- unknown
- Importer
- LLM Stats
- Reported scores
- 23
Source provenance
- Original evidence https://api.zeroeval.com/leaderboard/benchmarks/polymath?top_n=500
PolyMATH results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Qwen3 VL 30B A3B Instruct | Qwen | 0.443 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.517 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.405 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.52 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.288 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.446 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.304 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.475 | 2025-09-22 | evidence |
| Qwen3-235B-A22B-Instruct-2507 | Qwen | 0.502 | 2025-07-22 | evidence |
| Qwen3-235B-A22B-Thinking-2507 | Qwen | 0.601 | 2025-07-25 | evidence |
| Qwen3-Next-80B-A3B-Instruct | Qwen | 0.459 | 2025-09-10 | evidence |
| Qwen3-Next-80B-A3B-Thinking | Qwen | 0.563 | 2025-09-10 | evidence |
| Qwen3.5-0.8B | Qwen | 0.082 | 2026-03-02 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.689 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.712 | 2026-02-24 | evidence |
| Qwen3.5-2B | Qwen | 0.261 | 2026-03-02 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.644 | 2026-02-24 | evidence |
| Qwen3.5-397B-A17B | Qwen | 0.733 | 2026-02-16 | evidence |
| Qwen3.5-4B | Qwen | 0.511 | 2026-03-02 | evidence |
| Qwen3.5-9B | Qwen | 0.573 | 2026-03-02 | evidence |
| Qwen3.6 Plus | Qwen | 0.774 | 2026-04-02 | evidence |
| Qwen3.7 Max | Qwen | 0.865 | 2026-05-19 | evidence |
| Qwen3.7-Plus | Qwen | 0.84 | 2026-05-31 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.