MATH 500 (Vals AI) leaderboard
- GPT-5 leads
- 96.0%
- models measured
- 25
- lead over #2
- +0.6 pts
- median
- 89.0%
Leader on MATH 500 (Vals AI) · read October 21
On October 21, 2026, 29 of our catalog's models had a MATH 500 (Vals AI) result, and OpenAI GPT-5 led with 96.0%, ahead of Anthropic Claude Opus 4.1 at 95.4%.
Five hundred competition math problems from the MATH dataset. Share solved, as run by Vals AI. An archived board.
Role in the index
MATH 500 (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Latency, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | GPT-5OpenAI | 96.0% | 32 s | 38.8 | $10.00 | |
| 2 | Claude Opus 4.1Anthropic | 95.4% | 46 s | Not benchmarked | $75.00 | |
| 3 | GPT-5 MiniOpenAI | 94.8% | 12 s | 34.2 | $2.00 | |
| 4 | gpt-oss-120bOpenAI | 94.8% | 16 s | Not benchmarked | $0.50 | |
| 5 | o3OpenAI | 94.6% | 17 s | Not benchmarked | $8.00 | |
| 6 | Kimi K2 0711Moonshot AI | 94.2% | 37 s | Not benchmarked | $2.30 | |
| 7 | gpt-oss-20bOpenAI | 94.2% | 31 s | Not benchmarked | $0.15 | |
| 8 | o4 MiniOpenAI | 94.2% | 13 s | Not benchmarked | $4.40 | |
| 9 | GLM 4.5Z.ai | 94.0% | 1m 25s | Not benchmarked | $2.20 | |
| 10 | GPT-5 NanoOpenAI | 93.8% | 22 s | Not benchmarked | $0.40 | |
| 11 | R1RetiredDeepSeek | 92.2% | 2m 36s | Not benchmarked | $2.50 | |
| 12 | o3 MiniOpenAI | 91.8% | 14 s | Not benchmarked | $4.40 | |
| 13 | o1OpenAI | 90.4% | 24 s | Not benchmarked | $60.00 | |
| 14 | MiniMax M2.1MiniMax | 89.0% | 60 s | Not benchmarked | $1.20 | |
| 15 | DeepSeek V3 0324DeepSeek | 88.6% | 13 s | Not benchmarked | $1.07 | |
| 16 | Gemini 2.0 FlashRetiredGoogle | 88.0% | 3.4 s | Not benchmarked | $0.40 | |
| 17 | GPT-4.1 MiniOpenAI | 88.0% | 5.8 s | Not benchmarked | $1.60 | |
| 18 | GPT-4.1OpenAI | 87.2% | 13 s | Not benchmarked | $8.00 | |
| 19 | Mistral Medium 3.1Mistral AI | 87.0% | 14 s | Not benchmarked | $2.00 | |
| 20 | Llama 4 MaverickMeta Llama | 85.2% | 7.5 s | Not benchmarked | $1.00 | |
| 21 | Gemini 1.5 ProRetiredGoogle | 82.8% | 5.0 s | Not benchmarked | $5.00 | |
| 22 | GPT-4.1 NanoOpenAI | 80.2% | 3.4 s | Not benchmarked | $0.40 | |
| 23 | Llama 4 ScoutMeta Llama | 79.2% | 11 s | Not benchmarked | $0.50 | |
| 24 | Gemini 1.5 FlashRetiredGoogle | 78.8% | 2.6 s | Not benchmarked | $0.30 | |
| 25 | Command ACohere | 76.2% | 8.7 s | Not benchmarked | $10.00 | |
| 26 | GPT-4o (2024-08-06)OpenAI | 75.2% | 12 s | Not benchmarked | $10.00 | |
| 27 | Mistral LargeMistral AI | 74.4% | 9.9 s | Not benchmarked | $6.00 | |
| 28 | GPT-4o (2024-11-20)OpenAI | 74.0% | 13 s | Not benchmarked | $10.00 | |
| 29 | GPT-4o-miniOpenAI | 72.6% | 6.3 s | Not benchmarked | $0.60 |
Only models in our catalog are listed, and each appears once with its best published run. 136 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.