MathArena leaderboard
- GPT-6.1 Sol leads
- 84.0%
- models measured
- 69
- lead over #2
- +2.1 pts
- median
- 20.4%
Leader on MathArena · read October 22
On October 22, 2026, 70 of our catalog's models had a MathArena result, and OpenAI GPT-6.1 Sol led with 84.0%, ahead of OpenAI GPT-6 Astra at 81.9%.
MathArena tests models on recent math competitions and olympiads, most of them held after the models were trained. The number is expected performance: the average normalized score over all non-deprecated competitions, as MathArena publishes it on its home page.
Role in the index
MathArena is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Expected cost per problem | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | GPT-6.1 SolOpenAI | 84.0% | $0.7427 | 82.4 | $10.00 | |
| 2 | GPT-6 AstraOpenAI | 81.9% | $3.0659 | 85.1 | $50.00 | |
| 3 | GPT-6 SolOpenAI | 80.1% | $3.1343 | 76.0 | $10.00 | |
| 4 | Claude Opus 5.5Anthropic | 79.0% | $3.188 | 88.8 | $20.00 | |
| 5 | Claude Fable 5.1Anthropic | 78.1% | $17.8963 | 84.2 | $50.00 | |
| 6 | Claude Opus 5Anthropic | 70.4% | $17.6665 | 83.0 | $25.00 | |
| 7 | Gemini 4 ArgonGoogle | 69.3% | $9.4605 | 81.0 | $10.00 | |
| 8 | GPT-5.6 SolOpenAI | 62.9% | Not published | 73.5 | $20.00 | |
| 9 | GPT-5.5OpenAI | 61.3% | $7.4842 | 71.3 | $30.00 | |
| 10 | GPT-5.4 ProOpenAI | 52.5% | $67.7639 | Not benchmarked | $180.00 | |
| 11 | Qwen3.8 Max (0902)Qwen | 51.6% | $4.5041 | 62.4 | $6.00 | |
| 12 | Claude Fable 5Anthropic | 51.5% | $45.0023 | 79.3 | $50.00 | |
| 13 | Kimi K3Moonshot AI | 48.0% | $5.4639 | 62.3 | $15.00 | |
| 14 | Muse Spark 1.3Meta | 43.0% | $3.2496 | 57.4 | $4.25 | |
| 15 | Muse Spark 1.1Meta | 42.5% | $1.5253 | 70.3 | $4.25 | |
| 16 | DeepSeek V4.1 FlashDeepSeek | 42.3% | $0.6263 | 62.5 | $0.84 | |
| 17 | Grok 4.7xAI | 41.1% | $7.4378 | 66.7 | $6.00 | |
| 18 | Claude Opus 4.8Anthropic | 39.8% | $31.2477 | 65.0 | $25.00 | |
| 19 | GPT-5.4OpenAI | 37.2% | $5.00 | 64.4 | $15.00 | |
| 20 | Grok 4.5xAI | 34.1% | $2.8403 | 58.6 | $6.00 | |
| 21 | Gemini 3.8 FlashGoogle | 34.0% | $1.1003 | 61.6 | $3.75 | |
| 22 | GPT-5.2OpenAI | 33.7% | $4.6214 | 55.9 | $14.00 | |
| 23 | Muse Spark OpenMeta | 33.6% | $1.1397 | 47.4 | $4.25 | |
| 24 | Qwen3.8 2.4T A95BQwen | 31.6% | $2.7626 | Not benchmarked | $6.00 | |
| 25 | Gemini 3.1 Pro PreviewGoogle | 30.1% | $3.7638 | 32.3 | $12.00 | |
| 26 | Mistral Large 4Mistral AI | 26.7% | $2.6739 | 56.4 | $4.18 | |
| 27 | DeepSeek V4 Pro 0423DeepSeek | 26.2% | $3.7576 | 38.7 | $2.60 | |
| 28 | Gemini 3.7 FlashGoogle | 25.9% | $0.8517 | 58.0 | $3.75 | |
| 29 | DeepSeek V4 Flash 0423DeepSeek | 24.8% | $0.362 | Not benchmarked | $0.28 | |
| 30 | Gemini 3.5 FlashGoogle | 24.7% | $2.7455 | 39.8 | $9.00 | |
| 31 | Kimi K2.6Moonshot AI | 24.5% | $2.5378 | 52.1 | $4.00 | |
| 32 | Gemini 3.6 FlashGoogle | 23.9% | $1.7894 | 62.5 | $3.75 | |
| 33 | Claude Opus 4.6Anthropic | 23.6% | $14.0312 | 57.2 | $25.00 | |
| 34 | GLM 5.2Z.ai | 22.7% | $1.9272 | 61.1 | $4.40 | |
| 35 | Claude Opus 4.7Anthropic | 20.4% | $14.0137 | 67.5 | $25.00 | |
| 36 | GLM 5.1Z.ai | 20.1% | $1.9289 | 49.2 | $4.40 | |
| 37 | GLM 5Z.ai | 19.8% | $1.8093 | 41.7 | $3.20 | |
| 38 | Step 3.5 FlashStepFun | 19.4% | $0.3216 | Not benchmarked | $0.30 | |
| 39 | Qwen3.5 397B A17BQwen | 19.2% | $1.6172 | Not benchmarked | $2.34 | |
| 40 | Step 3.7 FlashStepFun | 19.1% | $0.3983 | Not benchmarked | $1.15 | |
| 41 | Kimi K2.5Moonshot AI | 17.9% | $1.4758 | 38.6 | $2.85 | |
| 42 | GPT-5.1OpenAI | 17.3% | $3.7282 | 40.4 | $10.00 | |
| 43 | GPT-5OpenAI | 17.1% | $3.3104 | 38.8 | $10.00 | |
| 44 | Nemotron 3 SuperNVIDIA | 16.6% | Not published | Not benchmarked | $0.45 | |
| 45 | GLM 4.6Z.ai | 16.3% | $0.8148 | 8.8 | $2.20 | |
| 46 | DeepSeek V3.2DeepSeek | 16.0% | $0.1555 | 30.0 | $0.88 | |
| 47 | Kimi K2 ThinkingMoonshot AI | 15.9% | $1.3555 | Not benchmarked | $2.50 | |
| 48 | Qwen3.5-27BQwen | 15.8% | $1.0391 | Not benchmarked | $1.56 | |
| 49 | gpt-oss-120bOpenAI | 15.1% | $0.2747 | Not benchmarked | $0.50 | |
| 50 | GPT-5 MiniOpenAI | 14.8% | $0.5852 | 34.2 | $2.00 | |
| 51 | o4 MiniOpenAI | 14.8% | $1.3122 | Not benchmarked | $4.40 | |
| 52 | Qwen3.5-35B-A3BQwen | 14.7% | $0.8697 | Not benchmarked | $1.30 | |
| 53 | DeepSeek V3.2 ExpDeepSeek | 14.4% | $0.1441 | Not benchmarked | $0.41 | |
| 54 | o3OpenAI | 14.2% | $2.1924 | Not benchmarked | $8.00 | |
| 55 | GLM 4.5Z.ai | 13.2% | $1.0438 | Not benchmarked | $2.20 | |
| 56 | Qwen3.5-9BQwen | 13.0% | $0.0936 | Not benchmarked | $0.15 | |
| 57 | R1 0528DeepSeek | 12.9% | $0.9534 | Not benchmarked | $2.18 | |
| 58 | GPT-5 NanoOpenAI | 12.8% | $0.278 | Not benchmarked | $0.40 | |
| 59 | gpt-oss-20bOpenAI | 12.8% | $0.1902 | Not benchmarked | $0.15 | |
| 60 | Gemini 2.5 ProGoogle | 12.7% | Not published | 19.1 | $10.00 | |
| 61 | Claude Sonnet 4.5Anthropic | 12.6% | $4.7451 | 39.5 | $15.00 | |
| 62 | GLM 4.5 AirZ.ai | 12.0% | $0.6085 | Not benchmarked | $1.10 | |
| 63 | o3 MiniOpenAI | 11.6% | $1.204 | Not benchmarked | $4.40 | |
| 64 | Gemini 2.5 FlashGoogle | 10.6% | $1.8229 | Not benchmarked | $2.50 | |
| 65 | GLM 4.5VZ.ai | 10.5% | $0.2432 | Not benchmarked | $1.80 | |
| 66 | o1OpenAI | 10.2% | $17.0826 | Not benchmarked | $60.00 | |
| 67 | Qwen3 30B A3BQwen | 9.4% | $0.1167 | Not benchmarked | $0.50 | |
| 68 | R1RetiredDeepSeek | 9.1% | $0.5544 | Not benchmarked | $2.50 | |
| 69 | DeepSeek V3 0324DeepSeek | 7.1% | $0.0908 | Not benchmarked | $1.07 | |
| 70 | GPT-4oOpenAI | 3.6% | $0.1835 | Not benchmarked | $10.00 |
Only models in our catalog are listed, and each appears once with its best published run. 70 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Expected cost is the average cost of one problem in US dollars, as MathArena reports it for the same run.