AA Math Index leaderboard
- GPT-5.2 leads
- 99.0
- models measured
- 72
- lead over #2
- +2.0
- median
- 62.1
Leader on AA Math Index · read October 22
On October 22, 2026, 73 of our catalog's models had a AA Math Index result, and OpenAI GPT-5.2 led with 99.0, ahead of Google Gemini 3 Flash Preview at 97.0.
The math part of the Artificial Analysis evaluations, combined into one score from 0 to 100. Run by Artificial Analysis on its own harness.
Role in the index
AA Math Index is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-5.2OpenAI | 99.0 | 55.9 | $14.00 | |
| 2 | Gemini 3 Flash PreviewGoogle | 97.0 | 43.7 | $3.00 | |
| 3 | GPT-5.1-CodexOpenAI | 95.7 | 15.3 | $10.00 | |
| 4 | GLM 4.7Z.ai | 95.0 | 55.1 | $2.20 | |
| 5 | Kimi K2 ThinkingMoonshot AI | 94.7 | Not benchmarked | $2.50 | |
| 6 | GPT-5OpenAI | 94.3 | 38.8 | $10.00 | |
| 7 | GPT-5.1OpenAI | 94.0 | 40.4 | $10.00 | |
| 8 | gpt-oss-120bOpenAI | 93.4 | Not benchmarked | $0.50 | |
| 9 | DeepSeek V3.2DeepSeek | 92.0 | 30.0 | $0.88 | |
| 10 | GPT-5.1-Codex-MiniOpenAI | 91.7 | Not benchmarked | $2.00 | |
| 11 | Claude Opus 4.5Anthropic | 91.3 | 44.0 | $25.00 | |
| 12 | Nemotron 3 Nano 30B A3BNVIDIA | 91.0 | Not benchmarked | $0.22 | |
| 13 | Qwen3 235B A22B Instruct 2507Qwen | 91.0 | Not benchmarked | $0.75 | |
| 14 | GPT-5 MiniOpenAI | 90.7 | 34.2 | $2.00 | |
| 15 | o4 MiniOpenAI | 90.7 | Not benchmarked | $4.40 | |
| 16 | DeepSeek V3.1 TerminusDeepSeek | 89.7 | Not benchmarked | $1.0142 | |
| 17 | gpt-oss-20bOpenAI | 89.3 | Not benchmarked | $0.15 | |
| 18 | o3OpenAI | 88.3 | Not benchmarked | $8.00 | |
| 19 | Qwen3 VL 235B A22B InstructQwen | 88.3 | Not benchmarked | $1.90 | |
| 20 | DeepSeek V3.2 ExpDeepSeek | 87.7 | Not benchmarked | $0.41 | |
| 21 | Gemini 2.5 ProGoogle | 87.7 | 19.1 | $10.00 | |
| 22 | GLM 4.6Z.ai | 86.0 | 8.8 | $2.20 | |
| 23 | GLM 4.6VZ.ai | 85.3 | Not benchmarked | $0.90 | |
| 24 | Qwen3 Next 80B A3B InstructQwen | 84.3 | Not benchmarked | $1.10 | |
| 25 | GPT-5 NanoOpenAI | 83.7 | Not benchmarked | $0.40 | |
| 26 | MiniMax M2.1MiniMax | 82.7 | Not benchmarked | $1.20 | |
| 27 | Qwen3 VL 30B A3B InstructQwen | 82.3 | Not benchmarked | $0.70 | |
| 28 | GLM 4.5 AirZ.ai | 80.7 | Not benchmarked | $1.10 | |
| 29 | MiniMax M2MiniMax | 78.3 | Not benchmarked | $1.02 | |
| 30 | R1 0528DeepSeek | 76.0 | Not benchmarked | $2.18 | |
| 31 | GLM 4.5Z.ai | 73.7 | Not benchmarked | $2.20 | |
| 32 | Gemini 2.5 FlashGoogle | 73.3 | Not benchmarked | $2.50 | |
| 33 | Qwen3 32BQwen | 73.0 | Not benchmarked | $0.40 | |
| 34 | GLM 4.5VZ.ai | 73.0 | Not benchmarked | $1.80 | |
| 35 | Qwen3 30B A3BQwen | 72.3 | Not benchmarked | $0.50 | |
| 36 | Qwen3 30B A3B Instruct 2507Qwen | 66.3 | Not benchmarked | $0.30 | |
| 37 | Qwen3 14BQwen | 58.0 | Not benchmarked | $0.23 | |
| 38 | Grok 3RetiredxAI | 58.0 | Not benchmarked | $15.00 | |
| 39 | Kimi K2 0905Moonshot AI | 57.3 | Not benchmarked | $2.50 | |
| 40 | Kimi K2 0711Moonshot AI | 57.0 | Not benchmarked | $2.30 | |
| 41 | Gemini 2.5 Flash LiteGoogle | 53.3 | Not benchmarked | $0.40 | |
| 42 | GPT-4.1 MiniOpenAI | 46.3 | Not benchmarked | $1.60 | |
| 43 | DeepSeek V3 0324DeepSeek | 41.0 | Not benchmarked | $1.07 | |
| 44 | Qwen3 Coder 480B A35BQwen | 39.3 | Not benchmarked | $1.50 | |
| 45 | Mistral Medium 3.1Mistral AI | 38.3 | Not benchmarked | $2.00 | |
| 46 | GPT-4.1OpenAI | 34.7 | Not benchmarked | $8.00 | |
| 47 | Reka Flash 3Reka | 33.7 | Not benchmarked | $0.20 | |
| 48 | Qwen3 VL 8B InstructQwen | 30.7 | Not benchmarked | $0.75 | |
| 49 | Mistral Medium 3Mistral AI | 30.3 | Not benchmarked | $2.00 | |
| 50 | Qwen3 Coder 30B A3B InstructQwen | 29.0 | Not benchmarked | $0.44 | |
| 51 | GPT-4.1 NanoOpenAI | 24.0 | Not benchmarked | $0.40 | |
| 52 | Gemma 3 27BGoogle | 20.7 | Not benchmarked | $0.30 | |
| 53 | Llama 4 MaverickMeta Llama | 19.3 | Not benchmarked | $1.00 | |
| 54 | Gemma 3 12BGoogle | 18.3 | Not benchmarked | $0.1625 | |
| 55 | Phi 4Microsoft | 18.0 | Not benchmarked | $0.14 | |
| 56 | Nova Premier 1.0Amazon | 17.3 | Not benchmarked | $12.50 | |
| 57 | GPT-4o-miniOpenAI | 14.7 | Not benchmarked | $0.60 | |
| 58 | Llama 4 ScoutMeta Llama | 14.0 | Not benchmarked | $0.50 | |
| 59 | Qwen2.5 72B InstructQwen | 14.0 | Not benchmarked | $0.40 | |
| 60 | Command ACohere | 13.0 | Not benchmarked | $10.00 | |
| 61 | Gemma 3 4BGoogle | 12.7 | Not benchmarked | $0.075 | |
| 62 | Llama 3.3 70B InstructMeta Llama | 7.7 | Not benchmarked | $0.72 | |
| 63 | Nova Lite 1.0Amazon | 7.0 | Not benchmarked | $0.24 | |
| 64 | Nova Pro 1.0Amazon | 7.0 | Not benchmarked | $3.20 | |
| 65 | Nova Micro 1.0Amazon | 6.0 | Not benchmarked | $0.14 | |
| 66 | Granite 4.0 MicroIBM Granite | 6.0 | Not benchmarked | $0.112 | |
| 67 | GPT-4oOpenAI | 6.0 | Not benchmarked | $10.00 | |
| 68 | Llama 3.1 8B InstructMeta Llama | 4.3 | Not benchmarked | $0.08 | |
| 69 | Mistral Small 3Mistral AI | 4.3 | Not benchmarked | $0.08 | |
| 70 | Llama 3.1 70B InstructMeta Llama | 4.0 | Not benchmarked | $0.56 | |
| 71 | Llama 3.2 3B InstructMeta Llama | 3.3 | Not benchmarked | $0.3325 | |
| 72 | Llama 3.2 1B InstructMeta Llama | 0.0 | Not benchmarked | $0.201 | |
| 73 | Mistral Large 2407Mistral AI | 0.0 | Not benchmarked | $6.00 |
Only models in our catalog are listed, and each appears once with its best published run. 199 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.