BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases7

MATH 500 leaderboard

As of 6 Nov 2026, the top score on MATH 500 is 96.4% by Gemini 3 Pro (Google), ahead of Grok 4 at 96.2%, across 57 model configurations with a published result (Vals AI, result dated 9 Jan 2026). MATH 500 is a 500-problem subset of the MATH competition dataset spanning algebra, counting, geometry and number theory, each with a short exact answer. MATH 500 is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 6 Nov 2026)

  • Gemini 3 Pro holds the top MATH 500 score at 96.4% (Vals AI, 9 Jan 2026).
  • Grok 4 is second on MATH 500 at 96.2% (Vals AI, 9 Jan 2026).
  • MATH 500: Five hundred competition problems in algebra, probability and trigonometry. Run by Vals AI (Vals AI).
  • 57 model configurations have a published MATH 500 result on BenchLeader (BenchLeader, 6 Nov 2026).
  • MATH 500 does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Five hundred competition problems in algebra, probability and trigonometry. Run by Vals AI.

Published by
Vals AI
Category
Maths
Index weight
Reference only
Models
57
Data as of
6 Nov 2026

Vals AI (vals.ai).

What the test looks like

MATH 500 is a 500-problem subset of the MATH competition dataset spanning algebra, counting, geometry and number theory, each with a short exact answer.

How it is scored

Accuracy, run by Vals AI at the stated reasoning effort.

What to keep in mind

Frontier models score above 95%, so it separates smaller models more than leaders, and the dataset is public.

Top 15
  1. 1Gemini 3 Pro (high)96.4%
  2. 2Grok 496.2%
  3. 3GPT-5 (high)96.0%
  4. 4Claude Opus 4.1 (thinking)95.4%
  5. 5Gemini 2.5 Pro95.2%
  6. 6GPT-5 mini (high)94.8%
  7. 7gpt-oss-120b94.8%
  8. 8o3 (high)94.6%
  9. 9Qwen3 235B A22B94.6%
  10. 10Grok 3 Mini Fast (high)94.2%
  11. 11Kimi K294.2%
  12. 12o4-mini (high)94.2%
  13. 13gpt-oss-20b94.2%
  14. 14GLM-4.594.0%
  15. 15Claude Sonnet 4 (thinking)93.8%
57 of 57
#
1Gemini 3 Prohigh96.4%60.1$0.0512026-01-09
2Grok 496.2%58.3$0.0372026-01-09
3GPT-5high96.0%57.9$0.0202026-01-09
4Claude Opus 4.1thinking95.4%57.8$0.2022026-01-09
5Gemini 2.5 Pro95.2%54.5$0.00792026-01-09
6GPT-5 minihigh94.8%54.6$0.00372026-01-09
7gpt-oss-120b94.8%45.5$0.00142026-01-09
8o3high94.6%53.8$0.0112026-01-09
9Qwen3 235B A22B94.6%50.0$0.00342026-01-09
10Grok 3 Mini Fasthigh94.2%53.6$0.00192026-01-09
11Kimi K294.2%51.0$0.00302026-01-09
12o4-minihigh94.2%50.3$0.00482026-01-09
13gpt-oss-20b94.2%44.0$0.00102026-01-09
14GLM-4.594.0%51.6$0.0132026-01-09
15Claude Sonnet 4thinking93.8%53.3$0.0682026-01-09
16GPT-5 nanohigh93.8%47.4$0.00172026-01-09
17Claude Opus 4.193.0%51.4$0.0462026-01-09
18DeepSeek R192.2%47.3$0.0112026-01-09
19Gemini 2.5 Flashthinking91.8%48.3$0.0112026-01-09
20o3-minihigh91.8%47.7$0.00772026-01-09
21Gemini 2.5 Flash91.6%52.7$0.00652026-01-09
22Claude 3.7 Sonnetthinking91.6%51.9$0.1252026-01-09
23Langston Nim Nvidia Llama 3.3 Nemotron Super 49B v1 42e84561thinking91.4%––2026-01-09
24Claude Opus 490.4%52.6$0.0452026-01-09
25o1high90.4%48.3$0.0952026-01-09
26Claude Sonnet 490.3%50.5$0.0102026-01-09
27Grok 389.8%52.5$0.0132026-01-09
28MiniMax-M2.189.0%55.0$0.00462026-01-09
29Gemini 2.0 Flash89.0%43.5$0.00022026-01-09
30GPT-4.1 minihigh88.0%46.3$0.00122026-01-09
31GPT-4.1high87.2%46.6$0.00942026-01-09
32Mistral Medium 387.0%44.3$0.00182026-01-09
33Llama 4 Maverick85.2%42.1$0.00062026-01-09
34Gemini 2.0 Flash Thinkingthinking84.6%45.3$0.00042026-01-09
35Gemini 1.5 Pro 00282.8%41.5$0.00222026-01-09
36DeepSeek V380.4%44.4$0.00062026-01-09
37GPT-4.1 nanohigh80.2%32.2$0.00032026-01-09
38Llama 4 Scout79.2%38.6$0.00042026-01-09
39Gemini 1.5 Flash 00278.8%38.2$0.00012026-01-09
40Grok 278.4%42.3$0.00632026-01-09
41Claude 3.7 Sonnet76.8%49.9$0.00692026-01-09
42Command A76.2%42.2$0.00602026-01-09
43GPT-4ohigh75.2%41.6$0.00562026-01-09
44Mistral Large 2.174.4%42.6$0.00332026-01-09
45Llama 3.3 70B73.4%40.0$0.00072026-01-09
46GPT-4o minihigh72.6%33.6$0.00032026-01-09
47Claude 3.5 Sonnet72.4%47.1$0.00502026-01-09
48Llama Meta Llama 3.1 405B Turbo71.4%–$0.00302026-01-09
49Langston Nim Nvidia Llama 3.3 Nemotron Super 49B v171.2%––2026-01-09
50Mistral Small 240270.6%30.2$0.00032026-01-09
51Grok 3 Mini Fastlow70.2%50.6$0.00092026-01-09
52Mistral Small 3.168.4%38.5$0.00022026-01-09
53Llama Meta Llama 3.1 70B Turbo65.0%–$0.00102026-01-09
54Claude 3.5 Haiku64.2%38.8$0.00152026-01-09
55Jamba Large 1.654.8%26.0$0.00402026-01-09
56Llama Meta Llama 3.1 8B Turbo44.4%–$0.00042026-01-09
57Jamba Mini 1.625.4%21.8$0.00032026-01-09

More maths benchmarks · All LLM benchmarks

Related benchmarks

Top head-to-heads on MATH 500

Cite as: BenchLeader, “MATH 500 leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/vals_math500.

MATH 500: questions

What does MATH 500 measure?
MATH 500 is a 500-problem subset of the MATH competition dataset spanning algebra, counting, geometry and number theory, each with a short exact answer. Scores are reported in percent of tasks solved; higher is better.
What is the top score on MATH 500?
As of 6 Nov 2026, the highest score on MATH 500 is 96.4% by Gemini 3 Pro (result dated 9 Jan 2026), ahead of Grok 4 at 96.2%. Scores come from Vals AI and are re-read daily.
Which AI model leads MATH 500?
Gemini 3 Pro leads MATH 500 with 96.4% as of 6 Nov 2026 (result dated 9 Jan 2026), ahead of Grok 4 at 96.2%.
How many models have MATH 500 results?
57 model configurations have a MATH 500 result on BenchLeader, all taken from Vals AI.
Who runs MATH 500 and how often is it updated?
MATH 500 is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does MATH 500 count toward the BenchLeader Index?
No. MATH 500 is shown for reference but left out of the composite index.