BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases8

HELM Capabilities mean

As of 29 Oct 2026, the top score on HELM Capabilities mean is 81.9% by GPT-5 mini (OpenAI), ahead of o4-mini at 81.2%, across 68 model configurations with a published result (HELM Capabilities, result dated 24 Nov 2025). The mean of HELM's five capability scenarios: MMLU-Pro, GPQA, IFEval, WildBench and Omni-MATH. HELM Capabilities mean is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 29 Oct 2026)

  • GPT-5 mini holds the top HELM Capabilities mean score at 81.9% (HELM Capabilities, 24 Nov 2025).
  • o4-mini is second on HELM Capabilities mean at 81.2% (HELM Capabilities, 24 Nov 2025).
  • HELM Capabilities mean: Mean of HELM's five capability scenarios. Shown for reference (HELM Capabilities).
  • 68 model configurations have a published HELM Capabilities mean result on BenchLeader (BenchLeader, 29 Oct 2026).
  • HELM Capabilities mean does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Mean of HELM's five capability scenarios. Shown for reference.

Published by
HELM Capabilities
Category
Composite
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

The mean of HELM's five capability scenarios: MMLU-Pro, GPQA, IFEval, WildBench and Omni-MATH.

How it is scored

Unweighted mean of the five scores, on a 0 to 100 scale. Shown for reference only.

What to keep in mind

Five scenarios is a narrow base for a composite, and HELM updates roughly monthly, so new models appear with a lag.

Top 15
  1. 1GPT-5 mini81.9%
  2. 2o4-mini81.2%
  3. 3o381.1%
  4. 4GPT-580.7%
  5. 5Gemini 3 Pro79.9%
  6. 6Qwen3 235B A22B 250779.8%
  7. 7Grok 478.5%
  8. 8Claude Opus 4 (thinking)78.0%
  9. 9gpt-oss-120b77.0%
  10. 10Kimi K276.8%
  11. 11Claude Sonnet 4 (thinking)76.6%
  12. 12Claude Sonnet 4.576.2%
  13. 13Claude Opus 475.7%
  14. 14GPT-5 nano74.8%
  15. 15Gemini 2.5 Pro74.5%
68 of 68
#
1GPT-5 mini81.9%56.42025-11-24
2o4-mini81.2%56.92025-11-24
3o381.1%61.02025-11-24
4GPT-580.7%57.72025-11-24
5Gemini 3 Pro79.9%60.22025-11-24
6Qwen3 235B A22B 250779.8%50.92025-11-24
7Grok 478.5%58.42025-11-24
8Claude Opus 4thinking78.0%55.22025-11-24
9gpt-oss-120b77.0%45.52025-11-24
10Kimi K276.8%51.02025-11-24
11Claude Sonnet 4thinking76.6%53.32025-11-24
12Claude Sonnet 4.576.2%53.72025-11-24
13Claude Opus 475.7%52.62025-11-24
14GPT-5 nano74.8%55.92025-11-24
15Gemini 2.5 Pro74.5%54.52025-11-24
16Claude Sonnet 473.3%50.52025-11-24
17Grok 372.7%52.52025-11-24
18GPT-4.172.7%48.52025-11-24
19Qwen3 235B A22B72.6%50.02025-11-24
20GPT-4.1 mini72.6%44.82025-11-24
21Llama 4 Maverick71.8%42.22025-11-24
22Claude Haiku 4.571.7%47.72025-11-24
23Qwen3-Next 80B A3Bthinking70.0%52.32025-11-24
24DeepSeek R1 052869.9%49.42025-11-24
25Palmyra X569.6%–2025-11-24
26Grok 3 mini67.9%52.32025-11-24
27Gemini 2.0 Flash67.9%43.52025-11-24
28Claude 3.7 Sonnet67.4%49.92025-11-24
29gpt-oss-20b67.4%44.12025-11-24
30GLM-4.5-Air67.0%46.82025-11-24
31DeepSeek V366.5%44.52025-11-24
32Gemini 1.5 Pro 00265.7%41.52025-11-24
33GPT-5.165.6%56.82025-11-24
34Claude 3.5 Sonnet65.3%47.12025-11-24
35Llama 4 Scout64.4%38.62025-11-24
36Gemini 2.0 Flash-Lite64.2%48.42025-11-24
37Nova Premier63.7%46.12025-11-24
38GPT-4o63.4%43.32025-11-24
39Gemini 2.5 Flash62.6%52.72025-11-24
40Llama 3.1 405B61.8%41.92025-11-24
41GPT-4.1 nano61.6%37.02025-11-24
42Gemini 1.5 Flash 00260.9%38.32025-11-24
43Palmyra X 00460.9%–2025-11-24
44Qwen2.5 72B59.9%43.22025-11-24
45Mistral Large 2.159.8%42.62025-11-24
46Gemini 2.5 Flash-Lite59.1%48.32025-11-24
47Nova Pro59.1%40.92025-11-24
48Palmyra Fin57.7%–2025-11-24
49Granite 4.0 Small57.5%–2025-11-24
50Llama 3.1 70B57.4%40.92025-11-24
51GPT-4o mini56.5%35.42025-11-24
52Mistral Small 3.155.8%38.52025-11-24
53Nova Lite55.1%38.32025-11-24
54Claude 3.5 Haiku54.9%38.82025-11-24
55Qwen2.5 7B Instruct52.9%39.92025-11-24
56Nova Micro52.2%39.22025-11-24
57granite-4.0-micro48.6%36.42025-11-24
58Mixtral 8x22B Instruct47.8%37.52025-11-24
59Palmyra Med47.6%–2025-11-24
60Olmo 2 32B March 202547.5%–2025-11-24

More composite benchmarks · All LLM benchmarks

Related benchmarks

Top head-to-heads on HELM Capabilities mean

Cite as: BenchLeader, “HELM Capabilities mean leaderboard”, data as of 29 Oct 2026, https://www.benchleader.com/benchmarks/helm_mean.

HELM Capabilities mean: questions

What does HELM Capabilities mean measure?
The mean of HELM's five capability scenarios: MMLU-Pro, GPQA, IFEval, WildBench and Omni-MATH. Scores are reported in percent of tasks solved; higher is better.
What is the top score on HELM Capabilities mean?
As of 29 Oct 2026, the highest score on HELM Capabilities mean is 81.9% by GPT-5 mini (result dated 24 Nov 2025), ahead of o4-mini at 81.2%. Scores come from HELM Capabilities and are re-read daily.
Which AI model leads HELM Capabilities mean?
GPT-5 mini leads HELM Capabilities mean with 81.9% as of 29 Oct 2026, ahead of o4-mini at 81.2% (result dated 24 Nov 2025).
How many models have HELM Capabilities mean results?
68 model configurations have a HELM Capabilities mean result on BenchLeader, all taken from HELM Capabilities.
Who runs HELM Capabilities mean and how often is it updated?
HELM Capabilities mean is published by HELM Capabilities. BenchLeader re-reads the published results once a day and records the date each result was published.
Does HELM Capabilities mean count toward the BenchLeader Index?
No. HELM Capabilities mean is shown for reference but left out of the composite index.