BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases8

MMLU-Pro (HELM)

As of 29 Oct 2026, the top score on MMLU-Pro (HELM) is 90.3% by Gemini 3 Pro (Google), ahead of Claude Opus 4 at 87.5%, across 68 model configurations with a published result (HELM Capabilities, result dated 24 Nov 2025). MMLU-Pro is a harder successor to MMLU: about 12,000 questions across 14 subjects with ten answer options each and more reasoning-heavy items, removing the trivial and noisy questions of the original. MMLU-Pro (HELM) is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 29 Oct 2026)

  • Gemini 3 Pro holds the top MMLU-Pro (HELM) score at 90.3% (HELM Capabilities, 24 Nov 2025).
  • Claude Opus 4 is second on MMLU-Pro (HELM) at 87.5% (HELM Capabilities, 24 Nov 2025).
  • MMLU-Pro (HELM): Stanford CRFM's own MMLU-Pro run. Shown for reference; Vals AI's run is in the index (HELM Capabilities).
  • 68 model configurations have a published MMLU-Pro (HELM) result on BenchLeader (BenchLeader, 29 Oct 2026).
  • MMLU-Pro (HELM) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Stanford CRFM's own MMLU-Pro run. Shown for reference; Vals AI's run is in the index.

Published by
HELM Capabilities
Category
Knowledge
Index weight
Reference only
Models
68
Data as of
29 Oct 2026
Newest result
24 Nov 2025

The publisher has posted no new results on this board since 24 Nov 2025. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.

Stanford Center for Research on Foundation Models (crfm.stanford.edu), Apache-2.0 results.

What the test looks like

MMLU-Pro is a harder successor to MMLU: about 12,000 questions across 14 subjects with ten answer options each and more reasoning-heavy items, removing the trivial and noisy questions of the original.

How it is scored

Accuracy with chain-of-thought prompting.

What to keep in mind

Stanford CRFM's own run of MMLU-Pro in HELM's fixed harness. Shown for reference; Vals AI's run is counted in the index. See also the MMLU-Pro board.

Top 15
  1. 1Gemini 3 Pro90.3%
  2. 2Claude Opus 4 (thinking)87.5%
  3. 3Claude Sonnet 4.586.9%
  4. 4GPT-586.3%
  5. 5Gemini 2.5 Pro86.3%
  6. 6o385.9%
  7. 7Claude Opus 485.9%
  8. 8Grok 485.1%
  9. 9Qwen3 235B A22B 250784.4%
  10. 10Claude Sonnet 4 (thinking)84.3%
  11. 11Claude Sonnet 484.3%
  12. 12GPT-5 mini83.5%
  13. 13o4-mini82.0%
  14. 14Kimi K281.9%
  15. 15Qwen3 235B A22B81.7%
68 of 68
#
1Gemini 3 Pro90.3%60.22025-11-24
2Claude Opus 4thinking87.5%55.22025-11-24
3Claude Sonnet 4.586.9%53.72025-11-24
4GPT-586.3%57.72025-11-24
5Gemini 2.5 Pro86.3%54.52025-11-24
6o385.9%61.02025-11-24
7Claude Opus 485.9%52.62025-11-24
8Grok 485.1%58.42025-11-24
9Qwen3 235B A22B 250784.4%50.92025-11-24
10Claude Sonnet 4thinking84.3%53.32025-11-24
11Claude Sonnet 484.3%50.52025-11-24
12GPT-5 mini83.5%56.42025-11-24
13o4-mini82.0%56.92025-11-24
14Kimi K281.9%51.02025-11-24
15Qwen3 235B A22B81.7%50.02025-11-24
16GPT-4.181.1%48.52025-11-24
17Llama 4 Maverick81.0%42.22025-11-24
18Palmyra X580.4%–2025-11-24
19Grok 3 mini79.9%52.32025-11-24
20gpt-oss-120b79.5%45.52025-11-24
21DeepSeek R1 052879.3%49.42025-11-24
22Grok 378.8%52.52025-11-24
23Qwen3-Next 80B A3Bthinking78.6%52.32025-11-24
24Claude 3.7 Sonnet78.4%49.92025-11-24
25GPT-4.1 mini78.3%44.82025-11-24
26GPT-5 nano77.8%55.92025-11-24
27Claude Haiku 4.577.7%47.72025-11-24
28Claude 3.5 Sonnet77.7%47.12025-11-24
29GLM-4.5-Air76.2%46.82025-11-24
30Llama 4 Scout74.2%38.62025-11-24
31gpt-oss-20b74.0%44.12025-11-24
32Gemini 2.0 Flash73.7%43.52025-11-24
33Gemini 1.5 Pro 00273.7%41.52025-11-24
34Nova Premier72.6%46.12025-11-24
35DeepSeek V372.3%44.52025-11-24
36Llama 3.1 405B72.3%41.92025-11-24
37Gemini 2.0 Flash-Lite72.0%48.42025-11-24
38GPT-4o71.3%43.32025-11-24
39Gemini 1.5 Flash 00267.8%38.32025-11-24
40Nova Pro67.3%40.92025-11-24
41Palmyra X 00465.7%–2025-11-24
42Llama 3.1 70B65.3%40.92025-11-24
43Gemini 2.5 Flash63.9%52.72025-11-24
44Qwen2.5 72B63.1%43.22025-11-24
45Mistral Small 3.161.0%38.52025-11-24
46Claude 3.5 Haiku60.5%38.82025-11-24
47GPT-4o mini60.3%35.42025-11-24
48Nova Lite60.0%38.32025-11-24
49Mistral Large 2.159.9%42.62025-11-24
50Palmyra Fin59.1%–2025-11-24
51GPT-5.157.9%56.82025-11-24
52Granite 4.0 Small56.9%–2025-11-24
53GPT-4.1 nano55.0%37.02025-11-24
54Qwen2.5 7B Instruct53.9%39.92025-11-24
55Gemini 2.5 Flash-Lite53.7%48.32025-11-24
56Nova Micro51.1%39.22025-11-24
57Mixtral 8x22B Instruct46.0%37.52025-11-24
58Olmo 2 32B March 202541.4%–2025-11-24
59Palmyra Med41.1%–2025-11-24
60Llama 3.1 8B40.6%33.72025-11-24

More knowledge benchmarks · All LLM benchmarks

Related benchmarks

Cite as: BenchLeader, “MMLU-Pro (HELM) leaderboard”, data as of 29 Oct 2026, https://www.benchleader.com/benchmarks/helm_mmlu_pro.

MMLU-Pro (HELM): questions

What does MMLU-Pro (HELM) measure?
MMLU-Pro is a harder successor to MMLU: about 12,000 questions across 14 subjects with ten answer options each and more reasoning-heavy items, removing the trivial and noisy questions of the original. Scores are reported in percent of tasks solved; higher is better.
What is the top score on MMLU-Pro (HELM)?
As of 29 Oct 2026, the highest score on MMLU-Pro (HELM) is 90.3% by Gemini 3 Pro (result dated 24 Nov 2025), ahead of Claude Opus 4 at 87.5%. Scores come from HELM Capabilities and are re-read daily.
Which AI model leads MMLU-Pro (HELM)?
Gemini 3 Pro leads MMLU-Pro (HELM) with 90.3% as of 29 Oct 2026, ahead of Claude Opus 4 at 87.5% (result dated 24 Nov 2025).
How many models have MMLU-Pro (HELM) results?
68 model configurations have a MMLU-Pro (HELM) result on BenchLeader, all taken from HELM Capabilities.
Who runs MMLU-Pro (HELM) and how often is it updated?
MMLU-Pro (HELM) is published by HELM Capabilities. BenchLeader re-reads the published results once a day and records the date each result was published.
Does MMLU-Pro (HELM) count toward the BenchLeader Index?
No. MMLU-Pro (HELM) is shown for reference but left out of the composite index.