BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases7

MMLU leaderboard

As of 6 Nov 2026, the top score on MMLU is 88.1% by GPT-4o (OpenAI), ahead of Claude 3.5 Sonnet at 87.3%, across 120 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 20 Nov 2024). Fifty-seven-subject multiple-choice exam of world knowledge. MMLU is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 6 Nov 2026)

  • GPT-4o holds the top MMLU score at 88.1% (Epoch AI Benchmarking Hub, 20 Nov 2024).
  • Claude 3.5 Sonnet is second on MMLU at 87.3% (Epoch AI Benchmarking Hub, 22 Oct 2024).
  • MMLU: Fifty-seven-subject multiple-choice exam of world knowledge. A 2020 benchmark, saturated for current models; kept for history (Epoch AI Benchmarking Hub).
  • 120 model configurations have a published MMLU result on BenchLeader (BenchLeader, 6 Nov 2026).
  • MMLU does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Fifty-seven-subject multiple-choice exam of world knowledge. A 2020 benchmark, saturated for current models; kept for history.

Published by
Epoch AI Benchmarking Hub
Category
Knowledge
Index weight
Reference only
Models
120
Data as of
6 Nov 2026

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

What the test looks like

Fifty-seven-subject multiple-choice exam of world knowledge.

How it is scored

Accuracy, as compiled by Epoch AI from published results.

What to keep in mind

Saturated: frontier models score near the ceiling, so it separates only older or smaller models, and results are largely self-reported by developers.

Top 15
  1. 1GPT-4o88.1%
  2. 2Claude 3.5 Sonnet87.3%
  3. 3DeepSeek V387.2%
  4. 4Gemini 1.5 Pro 00286.9%
  5. 5GPT-486.4%
  6. 6Llama 3.3 70B86.3%
  7. 7Gemini 1.5 Pro 00185.9%
  8. 8Qwen2.5 72B85.3%
  9. 9Phi-484.8%
  10. 10Claude 3 Opus84.6%
  11. 11Llama 3.1 405B84.5%
  12. 12Gemini 1.5 Pro 001 Feb2482.7%
  13. 13Qwen2-72B82.4%
  14. 14GPT 4 061382.4%
  15. 15Nova Pro82.0%
120 of 120
#
1GPT-4o88.1%43.32024-11-20
2Claude 3.5 Sonnet87.3%47.12024-10-22
3DeepSeek V387.2%44.42024-12-26
4Gemini 1.5 Pro 00286.9%41.52024-09-24
5GPT-486.4%42.42023-03-14
6Llama 3.3 70B86.3%40.02024-12-06
7Gemini 1.5 Pro 00185.9%–2024-05-14
8Qwen2.5 72B85.3%43.22024-09-19
9Phi-484.8%39.42024-12-12
10Claude 3 Opus84.6%39.02024-02-29
11Llama 3.1 405B84.5%41.92024-07-23
12Gemini 1.5 Pro 001 Feb2482.7%–2024-02-15
13Qwen2-72B82.4%41.32024-06-07
14GPT 4 061382.4%39.52023-06-13
15Nova Pro82.0%40.92024-12-03
16GPT-4o mini81.8%35.42024-07-18
17GPT-4 Turbo81.3%41.22024-04-09
18Llama 3.2 90B Vision80.3%36.12024-09-24
19Llama 3.1 70B80.1%40.82024-07-23
20Mistral Large 280.0%40.92024-07-24
21Qwen2.5 14B Instruct79.9%–2024-09-19
22Gemini 2.0 Flash79.7%43.52024-12-11
23Llama 3-70B79.3%38.72024-04-18
24Yi-Large79.3%–2024-05-13
25Qwen2.5 Coder 32B Instruct79.1%42.72024-09-18
26Claude 2.078.5%–2023-07-11
27Deepseek v278.4%–2024-05-07
28Phi 3medium78.0%37.92024-04-23
29Gemini 1.5 Flash 00177.9%–2024-05-23
30Gemini 1.5 Flash77.8%45.12024-05-14
31Mixtral 8x22B Instruct77.8%37.52024-04-17
32Nova Lite77.0%38.32024-12-03
33Claude 1.377.0%–2023-04-18
34Yi-34B76.3%34.92023-11-02
35Claude 3 Sonnet75.9%38.02024-02-29
36Gemma 2 27B75.7%40.82024-06-24
37Phi 3 Small75.7%–2024-04-23
38Qwen2.5-Coder-14B75.2%–2024-09-18
39Qwen1.5-32B74.4%39.62024-02-04
40Claude 3.5 Haiku74.3%38.82024-10-22
41Gemini 1.5 Flash 00273.9%38.22024-09-24
42Claude 3 Haiku73.8%36.32024-03-07
43Claude 2.173.5%–2023-11-21
44Claude Instant 1.173.4%––
45Claude Instant 1.273.2%–2023-08-09
46Qwen2.5 7B Instruct72.9%39.82024-09-19
47Inflection-172.7%–2023-06-22
48Gemma 2 9B72.1%38.82024-06-24
49GPT 3.5 Turbo 110671.4%38.62023-11-06
50Nova Micro70.8%39.22024-12-03
51Mixtral 8x7B70.6%30.92023-12-11
52Falcon-180B70.6%–2023-09-06
53Gemini 1.0 Pro 00170.0%–2023-12-13
54text-davinci-00270.0%–2022-03-15
55Llama 2-70B69.9%33.62023-07-18
56Command R+69.4%33.82024-08-30
57PaLM 540B69.3%–2022-04-04
58GPT-3.5 Turbo (older v0613)68.9%–2023-06-13
59Mistral Large 24.0268.8%40.62024-02-26
60Phi 3 Mini68.8%36.92024-04-23

More knowledge benchmarks · All LLM benchmarks

Related benchmarks

Cite as: BenchLeader, “MMLU leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/mmlu.

MMLU: questions

What does MMLU measure?
Fifty-seven-subject multiple-choice exam of world knowledge. Scores are reported in percent of tasks solved; higher is better.
What is the top score on MMLU?
As of 6 Nov 2026, the highest score on MMLU is 88.1% by GPT-4o (result dated 20 Nov 2024), ahead of Claude 3.5 Sonnet at 87.3%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
Which AI model leads MMLU?
GPT-4o leads MMLU with 88.1% as of 6 Nov 2026 (result dated 20 Nov 2024), ahead of Claude 3.5 Sonnet at 87.3%.
How many models have MMLU results?
120 model configurations have a MMLU result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
Who runs MMLU and how often is it updated?
MMLU is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
Does MMLU count toward the BenchLeader Index?
No. MMLU is shown for reference but left out of the composite index.