BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases8

Omni-MATH (HELM)

As of 29 Oct 2026, the top score on Omni-MATH (HELM) is 72.2% by GPT-5 mini (OpenAI), ahead of o4-mini at 72.0%, across 68 model configurations with a published result (HELM Capabilities, result dated 24 Nov 2025). Omni-MATH is a set of olympiad-level mathematics problems across many topics and difficulty tiers, with answers checked against the reference. Omni-MATH (HELM) counts toward the BenchLeader maths score.

Key facts (as of 29 Oct 2026)

  • GPT-5 mini holds the top Omni-MATH (HELM) score at 72.2% (HELM Capabilities, 24 Nov 2025).
  • o4-mini is second on Omni-MATH (HELM) at 72.0% (HELM Capabilities, 24 Nov 2025).
  • Omni-MATH (HELM): Olympiad-level maths problems. Run by Stanford CRFM's HELM (HELM Capabilities).
  • 68 model configurations have a published Omni-MATH (HELM) result on BenchLeader (BenchLeader, 29 Oct 2026).
  • Omni-MATH (HELM) feeds the maths category of the BenchLeader Index (BenchLeader methodology).

Olympiad-level maths problems. Run by Stanford CRFM's HELM.

Published by
HELM Capabilities
Category
Maths
Index weight
1.0
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

Omni-MATH is a set of olympiad-level mathematics problems across many topics and difficulty tiers, with answers checked against the reference.

How it is scored

Accuracy on the problems, run by Stanford CRFM's HELM with the same prompt for every model.

What to keep in mind

Answer checking for olympiad problems is imperfect, and HELM runs models without extended reasoning budgets unless the listing says so, so scores sit below what labs report with heavy test-time compute.

Top 15
  1. 1GPT-5 mini72.2%
  2. 2o4-mini72.0%
  3. 3Qwen3 235B A22B 250771.8%
  4. 4o371.4%
  5. 5gpt-oss-120b68.8%
  6. 6Kimi K265.4%
  7. 7GPT-564.7%
  8. 8Claude Opus 4 (thinking)61.6%
  9. 9Grok 460.3%
  10. 10Claude Sonnet 4 (thinking)60.2%
  11. 11gpt-oss-20b56.5%
  12. 12Claude Haiku 4.556.1%
  13. 13Gemini 3 Pro55.6%
  14. 14Claude Sonnet 4.555.3%
  15. 15Qwen3 235B A22B54.8%
68 of 68
#
1GPT-5 mini72.2%56.42025-11-24
2o4-mini72.0%56.92025-11-24
3Qwen3 235B A22B 250771.8%50.92025-11-24
4o371.4%61.02025-11-24
5gpt-oss-120b68.8%45.52025-11-24
6Kimi K265.4%51.02025-11-24
7GPT-564.7%57.72025-11-24
8Claude Opus 4thinking61.6%55.22025-11-24
9Grok 460.3%58.42025-11-24
10Claude Sonnet 4thinking60.2%53.32025-11-24
11gpt-oss-20b56.5%44.12025-11-24
12Claude Haiku 4.556.1%47.72025-11-24
13Gemini 3 Pro55.6%60.22025-11-24
14Claude Sonnet 4.555.3%53.72025-11-24
15Qwen3 235B A22B54.8%50.02025-11-24
16GPT-5 nano54.7%55.92025-11-24
17Claude Sonnet 451.3%50.52025-11-24
18Claude Opus 451.1%52.62025-11-24
19GPT-4.1 mini49.1%44.82025-11-24
20Gemini 2.5 Flash-Lite48.0%48.32025-11-24
21GPT-4.147.1%48.52025-11-24
22Qwen3-Next 80B A3Bthinking46.7%52.32025-11-24
23GPT-5.146.4%56.82025-11-24
24Grok 346.4%52.52025-11-24
25Gemini 2.0 Flash45.9%43.52025-11-24
26DeepSeek R1 052842.4%49.42025-11-24
27Llama 4 Maverick42.2%42.22025-11-24
28Gemini 2.5 Pro41.6%54.52025-11-24
29Palmyra X541.5%–2025-11-24
30DeepSeek V340.3%44.52025-11-24
31GLM-4.5-Air39.1%46.82025-11-24
32Gemini 2.5 Flash38.4%52.72025-11-24
33Gemini 2.0 Flash-Lite37.4%48.42025-11-24
34Llama 4 Scout37.3%38.62025-11-24
35GPT-4.1 nano36.7%37.02025-11-24
36Gemini 1.5 Pro 00236.4%41.52025-11-24
37Nova Premier35.0%46.12025-11-24
38Claude 3.7 Sonnet33.0%49.92025-11-24
39Qwen2.5 72B33.0%43.22025-11-24
40Palmyra X 00432.0%–2025-11-24
41Grok 3 mini31.8%52.32025-11-24
42Gemini 1.5 Flash 00230.5%38.32025-11-24
43Granite 4.0 Small29.6%–2025-11-24
44Palmyra Fin29.5%–2025-11-24
45Qwen2.5 7B Instruct29.4%39.92025-11-24
46GPT-4o29.3%43.32025-11-24
47Mistral Large 2.128.1%42.62025-11-24
48GPT-4o mini28.0%35.42025-11-24
49Claude 3.5 Sonnet27.6%47.12025-11-24
50Llama 3.1 405B24.9%41.92025-11-24
51Mistral Small 3.124.8%38.52025-11-24
52Nova Pro24.2%40.92025-11-24
53Nova Lite23.3%38.32025-11-24
54Claude 3.5 Haiku22.4%38.82025-11-24
55Nova Micro21.4%39.22025-11-24
56Llama 3.1 70B21.0%40.92025-11-24
57granite-4.0-micro20.9%36.42025-11-24
58Granite 3.3 8B17.6%–2025-11-24
59Mixtral 8x22B Instruct16.3%37.52025-11-24
60Olmo 2 32B March 202516.1%–2025-11-24

More maths benchmarks · All LLM benchmarks

Related benchmarks

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

Omni-MATH (HELM): questions

What does Omni-MATH (HELM) measure?
Omni-MATH is a set of olympiad-level mathematics problems across many topics and difficulty tiers, with answers checked against the reference. Scores are reported in percent of tasks solved; higher is better.
What is the top score on Omni-MATH (HELM)?
As of 29 Oct 2026, the highest score on Omni-MATH (HELM) is 72.2% by GPT-5 mini (result dated 24 Nov 2025), ahead of o4-mini at 72.0%. Scores come from HELM Capabilities and are re-read daily.
Which AI model leads Omni-MATH (HELM)?
GPT-5 mini leads Omni-MATH (HELM) with 72.2% as of 29 Oct 2026, ahead of o4-mini at 72.0% (result dated 24 Nov 2025).
How many models have Omni-MATH (HELM) results?
68 model configurations have a Omni-MATH (HELM) result on BenchLeader, all taken from HELM Capabilities.
Who runs Omni-MATH (HELM) and how often is it updated?
Omni-MATH (HELM) is published by HELM Capabilities. BenchLeader re-reads the published results once a day and records the date each result was published.
Does Omni-MATH (HELM) count toward the BenchLeader Index?
Yes. Omni-MATH (HELM) contributes to the maths category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.