MMLU Pro (Artificial Analysis run) leaderboard
- Claude Opus 4.5 leads
- 89.5%
- models measured
- 86
- lead over #2
- +0.5 pts
- median
- 78.0%
Leader on MMLU Pro (Artificial Analysis run) · read October 21
On October 21, 2026, 89 of our catalog's models had a MMLU Pro (Artificial Analysis run) result, and Anthropic Claude Opus 4.5 led with 89.5%, ahead of Google Gemini 3 Flash Preview at 89.0%.
Harder multiple-choice questions across fourteen academic and professional subjects. Share answered correctly. Run by Artificial Analysis on its own harness.
Role in the index
MMLU Pro (Artificial Analysis run) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.5Anthropic | 89.5% | 44.0 | $25.00 | |
| 2 | Gemini 3 Flash PreviewGoogle | 89.0% | 43.7 | $3.00 | |
| 3 | MiniMax M2.1MiniMax | 87.5% | Not benchmarked | $1.20 | |
| 4 | GPT-5.2OpenAI | 87.4% | 55.9 | $14.00 | |
| 5 | GPT-5OpenAI | 87.1% | 38.8 | $10.00 | |
| 6 | GPT-5.1OpenAI | 87.0% | 40.4 | $10.00 | |
| 7 | DeepSeek V3.2DeepSeek | 86.2% | 30.0 | $0.88 | |
| 8 | Gemini 2.5 ProGoogle | 86.2% | 19.1 | $10.00 | |
| 9 | GPT-5.1-CodexOpenAI | 86.0% | 15.3 | $10.00 | |
| 10 | Gemini 2.5 Pro Preview 06-05Google | 85.8% | Not benchmarked | $10.00 | |
| 11 | GLM 4.7Z.ai | 85.6% | 55.1 | $2.20 | |
| 12 | o3OpenAI | 85.3% | Not benchmarked | $8.00 | |
| 13 | DeepSeek V3.1 TerminusDeepSeek | 85.1% | Not benchmarked | $1.0142 | |
| 14 | DeepSeek V3.2 ExpDeepSeek | 85.0% | Not benchmarked | $0.41 | |
| 15 | R1 0528DeepSeek | 84.9% | Not benchmarked | $2.18 | |
| 16 | Kimi K2 ThinkingMoonshot AI | 84.8% | Not benchmarked | $2.50 | |
| 17 | Qwen3 235B A22B Instruct 2507Qwen | 84.3% | Not benchmarked | $0.75 | |
| 18 | o1OpenAI | 84.1% | Not benchmarked | $60.00 | |
| 19 | GPT-5 MiniOpenAI | 83.7% | 34.2 | $2.00 | |
| 20 | Qwen3 VL 235B A22B InstructQwen | 83.6% | Not benchmarked | $1.90 | |
| 21 | GLM 4.5Z.ai | 83.5% | Not benchmarked | $2.20 | |
| 22 | Gemini 2.5 FlashGoogle | 83.2% | Not benchmarked | $2.50 | |
| 23 | o4 MiniOpenAI | 83.2% | Not benchmarked | $4.40 | |
| 24 | GLM 4.6Z.ai | 82.9% | 8.8 | $2.20 | |
| 25 | Kimi K2 0711Moonshot AI | 82.4% | Not benchmarked | $2.30 | |
| 26 | Qwen3 Next 80B A3B InstructQwen | 82.4% | Not benchmarked | $1.10 | |
| 27 | MiniMax M2MiniMax | 82.0% | Not benchmarked | $1.02 | |
| 28 | GPT-5.1-Codex-MiniOpenAI | 82.0% | Not benchmarked | $2.00 | |
| 29 | DeepSeek V3 0324DeepSeek | 81.9% | Not benchmarked | $1.07 | |
| 30 | Kimi K2 0905Moonshot AI | 81.9% | Not benchmarked | $2.50 | |
| 31 | GLM 4.5 AirZ.ai | 81.5% | Not benchmarked | $1.10 | |
| 32 | Llama 4 MaverickMeta Llama | 80.9% | Not benchmarked | $1.00 | |
| 33 | gpt-oss-120bOpenAI | 80.8% | Not benchmarked | $0.50 | |
| 34 | Qwen3 VL 30B A3B InstructQwen | 80.7% | Not benchmarked | $0.70 | |
| 35 | GPT-4.1OpenAI | 80.6% | Not benchmarked | $8.00 | |
| 36 | Qwen3 30B A3B Instruct 2507Qwen | 80.5% | Not benchmarked | $0.30 | |
| 37 | o3 MiniOpenAI | 80.2% | Not benchmarked | $4.40 | |
| 38 | Grok 3RetiredxAI | 79.9% | Not benchmarked | $15.00 | |
| 39 | GLM 4.6VZ.ai | 79.9% | Not benchmarked | $0.90 | |
| 40 | Qwen3 32BQwen | 79.8% | Not benchmarked | $0.40 | |
| 41 | Nemotron 3 Nano 30B A3BNVIDIA | 79.4% | Not benchmarked | $0.22 | |
| 42 | Qwen3 Coder 480B A35BQwen | 78.8% | Not benchmarked | $1.50 | |
| 43 | GLM 4.5VZ.ai | 78.8% | Not benchmarked | $1.80 | |
| 44 | GPT-4.1 MiniOpenAI | 78.1% | Not benchmarked | $1.60 | |
| 45 | GPT-5 NanoOpenAI | 78.0% | Not benchmarked | $0.40 | |
| 46 | Qwen3 30B A3BQwen | 77.7% | Not benchmarked | $0.50 | |
| 47 | Qwen3 14BQwen | 77.4% | Not benchmarked | $0.23 | |
| 48 | Mistral Medium 3Mistral AI | 76.0% | Not benchmarked | $2.00 | |
| 49 | Gemini 2.5 Flash LiteGoogle | 75.9% | Not benchmarked | $0.40 | |
| 50 | Sonar ProPerplexity | 75.5% | Not benchmarked | $15.00 | |
| 51 | Llama 4 ScoutMeta Llama | 75.2% | Not benchmarked | $0.50 | |
| 52 | Qwen3 VL 8B InstructQwen | 74.9% | Not benchmarked | $0.75 | |
| 53 | GPT-4oOpenAI | 74.8% | Not benchmarked | $10.00 | |
| 54 | gpt-oss-20bOpenAI | 74.8% | Not benchmarked | $0.15 | |
| 55 | o1-miniRetiredOpenAI | 74.2% | Not benchmarked | $4.40 | |
| 56 | GPT-4o (2024-05-13)OpenAI | 74.0% | Not benchmarked | $15.00 | |
| 57 | Nova Premier 1.0Amazon | 73.3% | Not benchmarked | $12.50 | |
| 58 | Qwen2.5 72B InstructQwen | 72.0% | Not benchmarked | $0.40 | |
| 59 | Phi 4Microsoft | 71.4% | Not benchmarked | $0.14 | |
| 60 | Llama 3.3 70B InstructMeta Llama | 71.3% | Not benchmarked | $0.72 | |
| 61 | Command ACohere | 71.2% | Not benchmarked | $10.00 | |
| 62 | Grok 2RetiredxAI | 70.9% | Not benchmarked | $10.00 | |
| 63 | Qwen3 Coder 30B A3B InstructQwen | 70.6% | Not benchmarked | $0.44 | |
| 64 | GPT-4 TurboOpenAI | 69.4% | Not benchmarked | $30.00 | |
| 65 | Nova Pro 1.0Amazon | 69.1% | Not benchmarked | $3.20 | |
| 66 | SonarPerplexity | 68.9% | Not benchmarked | $1.00 | |
| 67 | Mistral Large 2407Mistral AI | 68.3% | Not benchmarked | $6.00 | |
| 68 | Mistral Medium 3.1Mistral AI | 68.3% | Not benchmarked | $2.00 | |
| 69 | Llama 3.1 70B InstructMeta Llama | 67.6% | Not benchmarked | $0.56 | |
| 70 | Gemma 3 27BGoogle | 66.9% | Not benchmarked | $0.30 | |
| 71 | Reka Flash 3Reka | 66.9% | Not benchmarked | $0.20 | |
| 72 | GPT-4.1 NanoOpenAI | 65.7% | Not benchmarked | $0.40 | |
| 73 | Mistral Small 3Mistral AI | 65.2% | Not benchmarked | $0.08 | |
| 74 | GPT-4o-miniOpenAI | 64.8% | Not benchmarked | $0.60 | |
| 75 | Qwen2.5 Coder 32B InstructQwen | 63.5% | Not benchmarked | $1.00 | |
| 76 | SabaMistral AI | 61.1% | Not benchmarked | $0.60 | |
| 77 | Gemma 3 12BGoogle | 59.5% | Not benchmarked | $0.1625 | |
| 78 | Nova Lite 1.0Amazon | 59.0% | Not benchmarked | $0.24 | |
| 79 | N Hermes 3 70B InstructNous Research | 57.1% | Not benchmarked | $0.70 | |
| 80 | GPT-4OpenAI | 56.2% | Not benchmarked | $60.00 | |
| 81 | Mixtral 8x22B InstructMistral AI | 53.7% | Not benchmarked | $6.00 | |
| 82 | Nova Micro 1.0Amazon | 53.1% | Not benchmarked | $0.14 | |
| 83 | Mistral LargeMistral AI | 51.5% | Not benchmarked | $6.00 | |
| 84 | Llama 3.1 8B InstructMeta Llama | 47.6% | Not benchmarked | $0.08 | |
| 85 | GPT-3.5 TurboOpenAI | 46.2% | Not benchmarked | $1.50 | |
| 86 | Granite 4.0 MicroIBM Granite | 44.7% | Not benchmarked | $0.112 | |
| 87 | Gemma 3 4BGoogle | 41.7% | Not benchmarked | $0.075 | |
| 88 | Llama 3.2 3B InstructMeta Llama | 34.7% | Not benchmarked | $0.3325 | |
| 89 | Llama 3.2 1B InstructMeta Llama | 20.0% | Not benchmarked | $0.201 |
Only models in our catalog are listed, and each appears once with its best published run. 198 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out.