MATH-500 AI Model Rankings
| # | Model | MATH | Cost |
|---|---|---|---|
| 1 | GPT 5 OpenAI | 98.1% | $3.44/M |
| 2 | GPT 5 Mini OpenAI | 97.8% | $0.69/M |
| 3 | o4-mini (High) OpenAI | 97.8% | $1.93/M |
| 4 | o3 OpenAI | 97.8% | $3.50/M |
| 5 | Qwen3 Max Thinking Qwen | 97.1% | $1.56/M |
| 6 | DeepSeek R1 (0528) DeepSeek | 96.6% | $0.91/M |
| 7 | o3-mini (High) OpenAI | 96.5% | $1.93/M |
| 8 | Gemini 2.5 Pro (05-06) Google | 95.9% | $3.44/M |
| 9 | GPT 5 Nano OpenAI | 95.2% | $0.14/M |
| 10 | o1 OpenAI | 94.7% | $26.25/M |
| 11 | DeepSeek R1 DeepSeek | 93.1% | $1.15/M |
| 12 | DeepSeek R1 Distill Llama 70B DeepSeek | 89.9% | $0.80/M |
| 13 | GPT 4.1 Mini OpenAI | 87.3% | $0.70/M |
| 14 | Claude Haiku 4.5 Anthropic | 86.9% | $2.00/M |
| 15 | Claude Opus 4 Anthropic | 85.0% | $30.00/M |
| 16 | Claude Sonnet 4 Anthropic | 84.4% | $6.00/M |
| 17 | GPT 4.1 OpenAI | 83.0% | $3.50/M |
| 18 | DeepSeek V3 (0324) DeepSeek | 75.5% | $0.34/M |
| 19 | Gemma 3 27B Google | 74.0% | $0.10/M |
| 20 | GPT 4.1 Nano OpenAI | 70.0% | $0.17/M |
| 21 | Qwen3 235B A22B Qwen | 68.9% | $0.80/M |
| 22 | Qwen Max Qwen | 67.2% | — |
| 23 | Qwen Plus Qwen | 65.3% | $0.39/M |
| 24 | Phi 4 Microsoft | 64.9% | $0.08/M |
| 25 | DeepSeek V3 DeepSeek | 64.9% | $0.35/M |
| 26 | Qwen2.5 72B Qwen | 63.2% | $0.37/M |
| 27 | Claude 3.5 Sonnet Anthropic | 56.9% | — |
| 28 | Qwen Turbo Qwen | 56.2% | — |
| 29 | GPT 4o OpenAI | 53.3% | $4.38/M |
| 30 | GPT 4o Mini OpenAI | 52.6% | $0.26/M |
| 31 | Mistral Large 2 (Nov 2024) Mistral | 50.3% | — |
| 32 | Llama 3.1 405B Meta | 49.8% | — |
| 33 | GPT 4 Turbo OpenAI | 46.7% | $15.00/M |
| 34 | Claude 3.5 Haiku Anthropic | 46.4% | $1.60/M |
| 35 | Mistral Large 2 (Jul 2024) Mistral | 44.8% | $3.00/M |
| 36 | Llama 3.3 70B Meta | 41.6% | $0.15/M |
| 37 | Llama 3.1 70B Meta | 36.7% | $0.40/M |
| 38 | Gemma 2 27B Google | 27.9% | $0.65/M |
| 39 | WizardLM 2 8x22B Microsoft | 25.7% | $0.62/M |
| 40 | Llama 3.1 8B Meta | 22.9% | $0.02/M |
| 41 | Gemma 2 9B Google | 21.0% | — |
| 42 | Claude 3 Haiku Anthropic | 14.9% | $0.50/M |
| 43 | Mistral 7B v0.3 Mistral | 3.6% | — |
43 of 43 models, ranked by MATH-500