BlogLLM Benchmark Scores as Published, Side by Side
PolyMATH: Results as Published
Every result the catalog holds for PolyMATH: 18 models from 6 developers, read from 3 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Science and MathMaintainer Qwen Team, Alibaba Group (Yiming Wang and co-authors)Introduced 2025Data version 2026103001, as of 2026-10-30
PolyMath tests mathematical reasoning in 18 languages. The same problems are written in each language and divided into four difficulty levels, from school mathematics through olympiad and frontier problems, with 125 problems per language at each level. Translations are calibrated by language experts rather than left as raw model output. The paper scores results with a difficulty-weighted accuracy made for this set.
PolyMATH
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.2OpenAI, released 2025-12-11 | 62.5% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| GPT-5 miniOpenAI, released 2025-08-07 | 67.3% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| gpt-oss-120bOpenAI, released 2025-08-05 | 54% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Claude Opus 4.5Anthropic, released 2025-11-24 | 79% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| Gemini 3 ProGoogle, released 2025-11-18 | 81.6% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| Qwen3.5-122B-A10BAlibaba, released 2026-02-24 | 68.9% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3.5-27BAlibaba, released 2026-02-24 | 71.2% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3.5-35B-A3BAlibaba, released 2026-02-24 | 64.4% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3.5-397B-A17BAlibaba, released 2026-02-16 | 73.3% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| Qwen3-235B-A22BAlibaba, released 2025-04-29 | 60.1% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3-Max-ThinkingAlibaba | 64.7% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| Kimi K2.5Moonshot AI, released 2026-01-27 | 43.1% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
PolyMath (en)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-4.1OpenAI, released 2025-04-14 | 54% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | - |
| Claude Opus 4Anthropic, released 2025-05-22 | 49.8% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | Effort: no extended thinking |
| Claude Sonnet 4Anthropic, released 2025-05-22 | 52.8% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | Effort: no extended thinking |
| Gemini 2.5 FlashGoogle, released 2025-05-20 | 49.9% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | - |
| DeepSeek-V3-0324DeepSeek, released 2025-03-24 | 59.5% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | - |
| Qwen3-235B-A22BAlibaba, released 2025-04-29 | 51.9% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | Effort: non-thinking |
| Kimi K2 InstructMoonshot AI, released 2025-07-11 | 65.1% | Moonshot AI, Kimi K2 release pagechecked 2026-10-15 | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.