MMLU Leaderboard
Multitask academic knowledge across 57 subjects. · Prices checked November 17, 2026
MMLU tests multiple-choice knowledge across 57 subjects, from elementary mathematics to professional law. It was the standard general-capability benchmark for years and remains a useful breadth check.
Leaderboard
60 of 152 scoredLeaderboardScores as published by each benchmark. Prices checked November 17, 2026.
o3 leads on MMLU at 92.9%. The median across the 152 models we hold a score for is 80%. The cheapest model on this board is Phi-4 at $0.0910/M blended, scoring 84.8%.
| # | Model | Provider | Score | Input /1M | Output /1M | Blended 70/30 | Context |
|---|---|---|---|---|---|---|---|
| 1 | o3 | OpenAI | 92.9% | $2.00 | $8.00 | $3.80 | 200,000 |
| 2 | o1 | OpenAI | 91.8% | $15.00 | $60.00 | $28.50 | 200,000 |
| 3 | GPT-5 | OpenAI | 91% | $1.25 | $10.00 | $3.88 | 400,000 |
| 4 | DeepSeek-R1 | DeepSeek | 90.8% | $0.5500 | $2.19 | $1.04 | 64,000 |
| 5 | DeepSeek-R1 (Groq) | Groq | 90.8% | $0.7500 | $0.9900 | $0.8220 | 128,000 |
| 6 | DeepSeek-R1 (Together) | Together AI | 90.8% | $3.00 | $7.00 | $4.20 | 64,000 |
| 7 | DeepSeek-R1 (fw) | Fireworks AI | 90.8% | $3.00 | $8.00 | $4.50 | 64,000 |
| 8 | DeepSeek-R1 (DI) | DeepInfra | 90.8% | $0.5500 | $2.19 | $1.04 | 64,000 |
| 9 | DeepSeek-R1 (Cerebras) | Cerebras | 90.8% | $0.5500 | $0.9900 | $0.6820 | 64,000 |
| 10 | GPT-4.1 | OpenAI | 90.2% | $2.00 | $8.00 | $3.80 | 1,000,000 |
| 11 | o4-mini | OpenAI | 90% | $1.10 | $4.40 | $2.09 | 200,000 |
| 12 | Claude Sonnet 3.5 | Anthropic | 89.3% | $3.00 | $15.00 | $6.60 | 200,000 |
| 13 | GPT-4o | OpenAI | 88.7% | $2.50 | $10.00 | $4.75 | 128,000 |
| 14 | Llama 3.1 405B | Meta | 88.6% | $0.8000 | $0.8000 | $0.8000 | 128,000 |
| 15 | Llama 3.1 405B (Groq) | Groq | 88.6% | $2.99 | $2.99 | $2.99 | 128,000 |
| 16 | Llama 3.1 405B (Together) | Together AI | 88.6% | $3.50 | $3.50 | $3.50 | 128,000 |
| 17 | Llama 3.1 405B (fw) | Fireworks AI | 88.6% | $3.00 | $3.00 | $3.00 | 131,000 |
| 18 | Llama 3.1 405B (Rep) | Replicate | 88.6% | $9.50 | $9.50 | $9.50 | 128,000 |
| 19 | Llama 3.1 405B (DI) | DeepInfra | 88.6% | $0.8000 | $0.8000 | $0.8000 | 128,000 |
| 20 | Llama-3.1-405B (Cerebras) | Cerebras | 88.6% | $0.9900 | $0.9900 | $0.9900 | 128,000 |
| 21 | Llama 3.1 405B (Hyp) | Hyperbolic | 88.6% | $4.00 | $4.00 | $4.00 | 128,000 |
| 22 | DeepSeek-V3 | DeepSeek | 88.5% | $0.2574 | $1.03 | $0.4888 | 64,000 |
| 23 | MiniMax-Text-01 | MiniMax | 88.5% | $0.7000 | $2.80 | $1.33 | 1,000,000 |
| 24 | GPT-4.1 mini | OpenAI | 87.5% | $0.4000 | $1.60 | $0.7600 | 1,000,000 |
| 25 | Grok 2 | xAI | 87.5% | $2.00 | $10.00 | $4.40 | 131,000 |
| 26 | Grok 2 Vision | xAI | 87.5% | $2.00 | $10.00 | $4.40 | 32,000 |
| 27 | Nova Premier | Amazon | 87.4% | $2.50 | $12.50 | $5.50 | 300,000 |
| 28 | Kimi k1.5 | Moonshot AI | 87.4% | $2.50 | $10.00 | $4.75 | 128,000 |
| 29 | Sonar | Perplexity | 87.1% | $1.00 | $1.00 | $1.00 | 127,000 |
| 30 | Qwen2.5-Max | Qwen | 87% | $1.60 | $6.40 | $3.04 | 32,000 |
| 31 | GLM-4-Plus | Zhipu AI | 86.8% | $7.00 | $7.00 | $7.00 | 128,000 |
| 32 | GLM-4V | Zhipu AI | 86.8% | $7.00 | $7.00 | $7.00 | 2,000 |
| 33 | GPT-4 Turbo | OpenAI | 86.7% | $10.00 | $30.00 | $16.00 | 128,000 |
| 34 | Qwen2.5-72B (Groq) | Groq | 86.1% | $0.7900 | $0.7900 | $0.7900 | 128,000 |
| 35 | Qwen2.5-72B (Together) | Together AI | 86.1% | $1.20 | $1.20 | $1.20 | 32,000 |
| 36 | Qwen2.5-72B (fw) | Fireworks AI | 86.1% | $0.9000 | $0.9000 | $0.9000 | 32,000 |
| 37 | Qwen2.5-72B (DI) | DeepInfra | 86.1% | $0.3500 | $0.4000 | $0.3650 | 128,000 |
| 38 | Qwen2.5-72B (Hyp) | Hyperbolic | 86.1% | $0.4000 | $0.4000 | $0.4000 | 128,000 |
| 39 | Qwen2.5-72B-Instruct | Qwen | 86.1% | $0.3600 | $0.4000 | $0.3720 | 128,000 |
| 40 | Gemini 1.5 Pro | 86% | $1.25 | $5.00 | $2.38 | 2,000,000 | |
| 41 | Llama 3.3 70B | Meta | 86% | $0.2300 | $0.4000 | $0.2810 | 128,000 |
| 42 | Llama 3.1 70B | Meta | 86% | $0.3500 | $0.4000 | $0.3650 | 128,000 |
| 43 | Llama 3.2 90B Vision | Meta | 86% | $0.8800 | $0.8800 | $0.8800 | 128,000 |
| 44 | Llama 3.3 70B (Groq) | Groq | 86% | $0.5900 | $0.7900 | $0.6500 | 128,000 |
| 45 | Llama 3.1 70B (Groq) | Groq | 86% | $0.5900 | $0.7900 | $0.6500 | 128,000 |
| 46 | Llama 3.3 70B (Together) | Together AI | 86% | $0.8800 | $0.8800 | $0.8800 | 128,000 |
| 47 | Llama 3.3 70B (DI) | DeepInfra | 86% | $0.2300 | $0.4000 | $0.2810 | 128,000 |
| 48 | Llama 3.1 70B (DI) | DeepInfra | 86% | $0.3500 | $0.4000 | $0.3650 | 128,000 |
| 49 | Llama-3.3-70B (Cerebras) | Cerebras | 86% | $0.5900 | $0.9900 | $0.7100 | 128,000 |
| 50 | Llama 3.1 70B (Hyp) | Hyperbolic | 86% | $0.4000 | $0.4000 | $0.4000 | 128,000 |
| 51 | Hermes-3-70B (Hyp) | Hyperbolic | 86% | $0.4000 | $0.4000 | $0.4000 | 128,000 |
| 52 | Llama 3.1 70B (Anyscale) | Anyscale | 86% | $1.00 | $1.00 | $1.00 | 128,000 |
| 53 | Llama 3.1 70B (Lepton) | Lepton AI | 86% | $0.8000 | $0.8000 | $0.8000 | 128,000 |
| 54 | Llama 3.1 70B (DB) | Databricks | 86% | $1.00 | $3.00 | $1.60 | 128,000 |
| 55 | o3-mini | OpenAI | 85.9% | $1.10 | $4.40 | $2.09 | 200,000 |
| 56 | Nova Pro | Amazon | 85.9% | $0.8000 | $3.20 | $1.52 | 300,000 |
| 57 | Llama 4 Maverick | Meta | 85.5% | $0.3500 | $1.00 | $0.5450 | 128,000 |
| 58 | Command A | Cohere | 85.5% | $2.50 | $10.00 | $4.75 | 256,000 |
| 59 | Pi (Inflection-2.5) | Inflection AI | 85.5% | $2.50 | $10.00 | $4.75 | 32,000 |
| 60 | Phi-4 | Microsoft | 84.8% | $0.0700 | $0.1400 | $0.0910 | 16,000 |
56 of the 60 scores shown link to the page they were read on (underlined; hover for publisher, variant and date read). The rest have no recorded source yet and are being worked through.
It is now close to saturated at the top: the strongest models cluster within a few points of each other, so small differences here mean much less than they did when the benchmark was new.
About these scores
- Multitask academic knowledge across 57 subjects.
- Scores are as published by the model’s provider or a public leaderboard. Provider-reported figures are self-reported and are not independently re-run by us.
- Where a hosted variant of an open-weights model has no score of its own, it inherits the base model’s, because the weights are the same.
- Prices beside each score are the provider’s published token rates. How the blended rate is calculated.
Showing the top 60 of 152 models we hold a MMLU score for.
Frequently asked questions
What is a good MMLU score?
Frontier models now sit in the high eighties and above, and the spread between them is only a few points. Because the benchmark is close to saturated at that level, MMLU is better used as a floor check (confirming broad competence) than as a way to separate the leading models from each other.
Is MMLU still a useful benchmark?
For breadth, yes; for ranking frontier models, less so. Its multiple-choice format rewards recall over reasoning, and top scores are bunched. GPQA Diamond and SWE-bench Verified separate the strongest models far more clearly.
Other leaderboards
Rankings that weigh price too
Leaderboards are recomputed on every deploy from the catalogue. Prices checked November 17, 2026.
