Category Score V3 leaderboard
LMSpeed Best Models for Math
Compare the best AI models for math across foundational, competition, proof, and tool-assisted mathematics benchmarks, with raw results shown beside the LMSpeed category estimate.
Methodology 3.0Methodology
Current answer
No model currently meets the formal ranking rules. The table has 89 Estimated models and 11 Provisional models. They are useful signals, but they are not formal ranks.
Available leaderboard data
- Models shown
- 100
- Formally ranked models
- 0
- Benchmark columns
- 13
- Dimensions with evidence
- 4/4
How to read the benchmark bars
Each bar compares models only within the same benchmark column. Bar lengths are relative to the models shown here; they are not Category Scores and cannot be compared across benchmark columns.
| Rank | Model | LMSpeed score | Foundational math | Competition math | Advanced proofs | Applied & tool-assisted math | Status | Evidence | Updated | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MATH-500undefined models | AA Math Indexundefined models | AIMEundefined models | HMMT Feb 2026undefined models | AIME26undefined models | HMMT Nov 2025undefined models | HMMT Feb 2025undefined models | AIME 2025undefined models | FrontierMath v2 (Tiers 1-3)undefined models | FrontierMath v2 (Tier 4)undefined models | FrontierMath (legacy)undefined models | IMOAnswerBenchundefined models | MMAnswerBenchundefined models | ||||||
| Estimated models — unranked89 | ||||||||||||||||||
| — | GLM-5.2Z.ai | 70.1±11.5 | — | — | — | 92.5 | 99.2 | 94.4 | — | — | — | — | — | — | 91.0 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | O4 MiniOpenAI | 63.9±11.8 | 98.9% | — | 94.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 235B A22B Instruct 2507Qwen | 62.9±11.8 | 98.4% | — | 94.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.7 MaxQwen | 62.2±12.2 | — | — | — | 97.1 | — | — | — | — | — | — | — | 90.0 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5OpenAI | 61.4±11.8 | 98.7% | — | 83.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | o3 Mini HighOpenAI | 61.3±11.8 | 98.5% | — | 86.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.5 Pro Preview 06-05Google | 60.7±11.8 | 98.0% | — | 87.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GLM-4.5Z.ai | 60.5±11.8 | 97.9% | — | 87.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek V4 Pro 0813DeepSeek | 58.9±12.2 | — | — | — | 95.2 | — | — | — | — | — | — | — | 89.8 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | MiniMax M1MiniMax | 58.9±11.8 | 97.2% | — | 81.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | O3OpenAI | 58.8±9.3 | 99.2% | — | 90.3% | — | — | — | — | — | 18.7 | 2.1 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Kimi K2.6MoonshotAI | 58.6±9.0 | — | — | — | 92.7 | 96.4 | — | — | — | 39.0 | 14.6 | — | — | 86.0 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | O3 MiniOpenAI | 58.5±11.8 | 97.3% | — | 77.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 | 58.3±11.8 | 97.5% | — | 72.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek V4 Flash 0731DeepSeek | 57.9±12.2 | — | — | — | 94.8 | — | — | — | — | — | — | — | 88.4 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Sonar Reasoning ProPerplexity | 57.4±11.8 | 95.7% | — | 79.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek R1DeepSeek | 57±11.8 | 96.6% | — | 68.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GLM-4.5 AirZ.ai | 56.9±11.8 | 96.5% | — | 67.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.5 ProGoogle | 56.1±9.3 | 96.7% | — | 88.7% | — | — | — | — | — | 14.1 | 4.2 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek R1 Distill QwenDeepSeek | 55.7±11.8 | 94.9% | — | 66.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.7 PlusQwen | 55.2±12.2 | — | — | — | 92.9 | — | — | — | — | — | — | — | 86.0 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | R1 Distill Llama 70BDeepSeek | 55±11.8 | 93.5% | — | 67.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | O1 MiniOpenAI | 54.9±11.8 | 94.4% | — | 60.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Opus 4Anthropic | 54.4±11.8 | 94.1% | — | 56.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Kimi K2MoonshotAI | 53.7±9.3 | 97.1% | — | 69.3% | — | — | — | — | — | 21.4 | 0.0 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.5 Flash LiteGoogle | 53.3±11.8 | 92.6% | — | 50.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Sonnet 4Anthropic | 52.9±11.8 | 93.4% | — | 40.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | O1OpenAI | 52.6±9.3 | 92.4% | — | 72.3% | — | — | — | — | — | 9.3 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Reka Flash 3Rekaai | 52.2±11.8 | 89.3% | — | 51.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.6 PlusQwen | 52.2±9.1 | — | — | — | 87.8 | 95.3 | 94.6 | 96.7 | — | 26.2 | 8.3 | — | — | 83.8 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Mistral Medium 3Mistral | 52.1±11.8 | 90.7% | — | 44.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.0 FlashGoogle | 52.1±11.8 | 93.0% | — | 33.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.0 ProGoogle | 52.1±11.8 | 92.3% | — | 36.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GLM-5.1Z.ai | 51.7±9.1 | — | — | — | 82.6 | 95.3 | 94.0 | — | — | 33.4 | 12.5 | — | — | 83.8 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GLM-5Z.ai | 51.1±9.1 | — | — | — | 86.4 | 95.8 | 96.9 | 97.5 | — | 16.4 | 2.1 | — | — | 82.5 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 235B A22BQwen | 51.1±11.8 | 90.2% | — | 32.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.5 FlashGoogle | 50.8±9.3 | 92.6% | — | 43.3% | — | — | — | — | — | 4.8 | 4.2 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 Coder 30B A3B InstructQwen | 50.5±11.8 | 89.3% | — | 29.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | SonarPerplexity | 50.3±11.8 | 81.7% | — | 48.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.0 Flash LiteGoogle | 50.1±11.8 | 87.3% | — | 30.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 32BQwen | 49.9±11.8 | 86.9% | — | 30.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4.1 MiniOpenAI | 49.9±9.3 | 92.5% | — | 43.0% | — | — | — | — | — | 4.5 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Opus 4.5Anthropic | 49.9±9.1 | — | — | — | 85.3 | 95.1 | 93.3 | 92.9 | — | 20.7 | 4.2 | — | — | 84.0 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 2.0 Flash Lite 001Google | 49.8±11.8 | 87.3% | — | 27.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 14BQwen | 49.8±11.8 | 87.1% | — | 28.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek V3 | 49.5±9.3 | 94.2% | — | 52.0% | — | — | — | — | — | 1.7 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 30B A3BQwen | 49.4±11.8 | 86.3% | — | 26.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Kimi K2.5MoonshotAI | 49.2±9.0 | — | — | — | 87.1 | 95.8 | 91.1 | 95.4 | 96.1 | 27.9 | 4.2 | — | — | 81.8 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4.1OpenAI | 49±9.3 | 91.3% | — | 43.7% | — | — | — | — | — | 5.5 | 0.0 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GLM-4.7Z.ai | 48.8±12.3 | — | — | — | — | — | — | — | 95.7 | 2.4 | 0.0 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3.7 SonnetAnthropic | 48.7±11.8 | 85.0% | — | 22.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3 8BQwen | 48.5±11.8 | 82.8% | — | 24.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Sonar ProPerplexity | 47.5±11.8 | 74.5% | — | 29.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Grok 3 | 47.3±9.3 | 87.0% | — | 33.0% | — | — | — | — | — | 3.8 | 0.0 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Phi 4Microsoft | 46.9±11.8 | 81.0% | — | 14.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Ling-3.0-flashInclusionai | 46.7±11.5 | — | — | — | 87.0 | 93.2 | — | — | — | — | — | — | 83.7 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Command ACohere | 46.2±11.8 | 81.9% | — | 9.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Grok-2 | 46.2±11.8 | 77.8% | — | 13.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4o (2024-08-06)OpenAI | 46.2±11.8 | 79.5% | — | 11.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Llama 4 MaverickMeta | 46.2±9.3 | 88.9% | — | 39.0% | — | — | — | — | — | 0.7 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4o MiniOpenAI | 46.1±11.8 | 78.9% | — | 11.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4o (2024-05-13)OpenAI | 46±11.8 | 79.1% | — | 11.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4 TurboOpenAI | 45.8±11.8 | 73.7% | — | 15.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek R1 Distill Qwen 1.5BDeepSeek | 45.5±11.8 | 68.7% | — | 17.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4.1 NanoOpenAI | 45.2±9.3 | 84.8% | — | 23.7% | — | — | — | — | — | 1.0 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Mistral SabaMistral | 44.7±11.8 | 67.7% | — | 13.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Mistral Large 2407Mistralai | 44.5±11.8 | 71.4% | — | 9.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Llama 4 ScoutMeta | 44.4±9.3 | 84.4% | — | 28.3% | — | — | — | — | — | 0.0 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Phi 4 Multimodal Instruct | 44.2±11.8 | 69.3% | — | 9.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 1.5 ProGoogle | 43.7±11.8 | 67.3% | — | 8.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Devstral SmallMistral AI | 43.4±11.8 | 68.4% | — | 6.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3.5 SonnetAnthropic | 43±9.3 | 69.5% | — | 9.7% | — | — | — | — | — | 2.1 | 0.0 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Gemini 1.5 FlashGoogle | 42.6±11.8 | 55.4% | — | 9.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3.5 HaikuAnthropic | 42.5±11.8 | 72.1% | — | 3.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Sonnet 4.5Anthropic | 42.5±12.3 | — | — | — | — | — | — | — | 87.0 | 13.5 | 4.2 | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-4oOpenAI | 42.2±9.3 | 79.7% | — | 10.3% | — | — | — | — | — | 0.3 | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Phi 4 Mini Instruct | 42±11.8 | 69.6% | — | 3.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3 OpusAnthropic | 41.5±11.8 | 64.1% | — | 3.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.5 | 41.4±11.5 | — | — | — | 87.9 | 93.3 | 92.7 | 94.8 | — | — | — | — | — | 80.9 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.6 27BQwen | 41.3±11.5 | — | — | — | 84.3 | 94.1 | 90.7 | 93.8 | — | — | — | — | — | 80.8 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Hermes 3 70B InstructNous | 39.6±11.8 | 53.8% | — | 2.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3 SonnetAnthropic | 39.6±11.8 | 41.4% | — | 4.7% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 2.1Anthropic | 38.5±11.8 | 37.4% | — | 3.3% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Mixtral 8x22B InstructMistral | 36.8±11.8 | 54.5% | — | 0.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Mistral LargeMistralai | 36.6±11.8 | 52.7% | — | 0.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude 3 HaikuAnthropic | 36.4±11.8 | 39.4% | — | 1.0% | — | — | — | — | — | — | — | — | — | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Qwen3.6 35B A3BQwen | 34.6±11.5 | — | — | — | 83.6 | 92.7 | 89.1 | 90.7 | — | — | — | — | — | 78.9 | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek V4 FlashDeepSeek | 32.6±12.2 | — | — | — | 40.8 | — | — | — | — | — | — | — | 41.9 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | DeepSeek V4 ProDeepSeek | 31.6±12.2 | — | — | — | 31.7 | — | — | — | — | — | — | — | 35.3 | — | Estimated | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| Provisional models — unranked11 | ||||||||||||||||||
| — | GPT-5.6 SolOpenAI | 67±16.1 | — | — | — | — | — | — | — | — | 89.0 | 83.0 | 89.0 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | InklingThinking Machines | 65.4±16.3 | — | — | — | — | 97.1 | — | — | — | — | — | — | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.6 TerraOpenAI | 64.5±16.1 | — | — | — | — | — | — | — | — | 84.9 | 68.3 | 84.9 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.6 LunaOpenAI | 62.4±16.1 | — | — | — | — | — | — | — | — | 78.6 | 58.5 | 78.6 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Opus 4.8Anthropic | 58±16.1 | — | — | — | — | — | — | — | — | 47.2 | 31.3 | — | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.4OpenAI | 57.6±16.1 | — | — | — | — | — | — | — | — | 47.6 | 27.1 | — | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.5 ProOpenAI | 57.2±16.1 | — | — | — | — | — | — | — | — | 51.0 | 39.6 | 52.4 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.5OpenAI | 56.9±16.1 | — | — | — | — | — | — | — | — | 51.7 | 35.4 | 51.7 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | GPT-5.4 ProOpenAI | 56.8±16.1 | — | — | — | — | — | — | — | — | 50.0 | 37.5 | 50.0 | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Opus 4.7Anthropic | 56.8±16.1 | — | — | — | — | — | — | — | — | 43.8 | 22.9 | — | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
| — | Claude Opus 4.6Anthropic | 56.5±16.1 | — | — | — | — | — | — | — | — | 40.7 | 22.9 | — | — | — | Provisional | undefined/4 dimensions · undefined families | Aug 28, 2026 |
What this leaderboard measures
Which AI model is better suited to math tasks?
This leaderboard covers foundational math, competition problems, advanced proofs, and tool-assisted mathematics. Different math tasks require different capabilities.
Four capability dimensions
The four dimensions come from the category blueprint. Available data may cover only some of them. A dimension without evidence is not presented as a verified capability.
Foundational math
2 benchmark columns currently provide evidence for this dimension.
- MATH-500
- AA Math Index
Competition math
6 benchmark columns currently provide evidence for this dimension.
- AIME
- HMMT Feb 2026
- AIME26
- HMMT Nov 2025
- HMMT Feb 2025
- AIME 2025
Advanced proofs
4 benchmark columns currently provide evidence for this dimension.
- FrontierMath v2 (Tiers 1-3)
- FrontierMath v2 (Tier 4)
- FrontierMath (legacy)
- IMOAnswerBench
Applied & tool-assisted math
1 benchmark columns currently provide evidence for this dimension.
- MMAnswerBench
Math tasks this page can help with
- Foundational problems in arithmetic, algebra, and common school mathematics.
- Competition problems and proofs that require a valid multi-step solution.
- Applied calculations where the model must set up formulas and check tool output.
How to choose a model with this leaderboard
- Step 1
Check the rating status first
Only Rated models receive a rank. Estimated and Provisional models do not have a formal position.
- Step 2
Review uncertainty and evidence
When scores are close, do not rely on rank alone. Check uncertainty, dimension coverage, and benchmark count.
- Step 3
Test the real task last
A leaderboard cannot replace your own test. Check quality, speed, price, context, and provider limits together.
Choose by problem type. Recalculate important values and check every proof step.
Rating status guide
Rated
Rated means the evidence and overlap rules are met. The model can receive a formal rank.
Estimated
Estimated means there is useful evidence, but it is not enough for a formal rank.
Provisional
Provisional means evidence is limited or dimension and benchmark-family coverage is below the estimated threshold. Use the result only as an early signal.
Benchmarks and evidence sources
Evidence source names and benchmark groups come from the currently available score data. One source may contribute several benchmarks.
Artificial Analysis
AA Math Index, AIME, and MATH-500
BenchLM
AIME 2025, AIME26, FrontierMath (legacy), FrontierMath v2 (Tier 4), FrontierMath v2 (Tiers 1-3), HMMT Feb 2025, HMMT Feb 2026, HMMT Nov 2025, IMOAnswerBench, and MMAnswerBench
How the math model ranking is built
LMSpeed combines eligible third-party benchmarks inside four fixed capability dimensions. Rated models meet the evidence and overlap requirements for a formal rank; Estimated and Provisional models remain visible without receiving a rank.
Read the Category Score methodologyLeaderboard limits
Category Scores use the third-party benchmarks currently included by LMSpeed. Tests can use different data, prompts, and scoring rules. The result is not permanent and cannot represent every real task. Test important choices with your own data and workflow.
Frequently asked questions
Which visible model has the highest formal rank now?
There is no formal number one now. The page has 89 Estimated models and 11 Provisional models. They do not have a formal rank and should not be called the winner.
Can I compare scores across different categories?
No. Each category uses different capability dimensions and evidence. A Category Score is comparable only inside the same leaderboard. Review the matching category for each task.
Are Estimated and Provisional models still useful?
They can help you find candidates, but their evidence is not complete enough for a formal rank. Review coverage and uncertainty, then test the model on a real task.
How often does the leaderboard update?
The leaderboard updates after a new completed score run is published. The current run date and methodology version appear above. LMSpeed does not promise a fixed daily or weekly schedule.
Is the number one model always best for me?
No. Your result also depends on speed, price, context length, tool support, region, and provider limits. Use the leaderboard to narrow the field, then run your own test.
Does the math ranking represent general reasoning?
No. The math leaderboard describes performance on math tasks. Logic, causality, and evidence verification belong in the reasoning leaderboard. Real work may also require tool use and stable answers.