BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

LiveBench Mathematics

Mathematics category of LiveBench 1.0, frozen at its final release (25 June 2026). LiveBench 2.0 dropped it.

As of 25 Oct 2026, Claude Opus 5.5 leads LiveBench Mathematics on BenchLeader with 97.1%, ahead of Claude Fable 5.1 at 97.0%, across 67 model configurations with a published result.

Published by
LiveBench
Category
Maths
Index weight
Reference only
Models
67
Data as of
25 Oct 2026
Newest result
25 Jun 2026

The publisher has posted no new results on this board since 25 Jun 2026. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.

LiveBench (livebench.ai), White et al. 2024.

What the test looks like

LiveBench's mathematics questions, drawn from recent competitions and freshly generated problems, with exact answers.

How it is scored

Percent correct, graded automatically, from LiveBench 1.0's final release (25 June 2026). LiveBench 2.0 dropped the category, so these scores no longer update.

What to keep in mind

Overlaps in style with AIME-type tests; strong here does not imply strong on proofs.

Top 15
  1. 1Claude Opus 5.5 (max)97.1%
  2. 2Claude Fable 5.1 (max)97.0%
  3. 3GPT-6.1 Sol (max)96.8%
  4. 4GPT-6 Astra (max)96.8%
  5. 5Claude Opus 5.5 (xhigh)96.8%
  6. 6Claude Sonnet 5.5 (xhigh)96.7%
  7. 7GPT-6.1 Sol (xhigh)96.5%
  8. 8Muse Wonder (max)96.5%
  9. 9GPT-6 Sol (max)96.4%
  10. 10GPT-5.6 Sol (max)96.2%
  11. 11Claude Sonnet 5.5 (max)96.1%
  12. 12Claude Fable 5 (max)96.0%
  13. 13Muse Spark 1.3 (xhigh)96.0%
  14. 14GPT-5.5 (xhigh)95.9%
  15. 15Claude Opus 5 (max)95.7%
67 of 67
#
1Claude Opus 5.5max97.1%72.62026-06-25
2Claude Fable 5.1max97.0%70.52026-06-25
3GPT-6.1 Solmax96.8%69.92026-06-25
4GPT-6 Astramax96.8%70.42026-06-25
5Claude Opus 5.5xhigh96.8%70.02026-06-25
6Claude Sonnet 5.5xhigh96.7%67.92026-06-25
7GPT-6.1 Solxhigh96.5%67.72026-06-25
8Muse Wondermax96.5%67.52026-06-25
9GPT-6 Solmax96.4%66.42026-06-25
10GPT-5.6 Solmax96.2%68.02026-06-25
11Claude Sonnet 5.5max96.1%69.42026-06-25
12Claude Fable 5max96.0%69.32026-06-25
13Muse Spark 1.3xhigh96.0%64.92026-06-25
14GPT-5.5xhigh95.9%66.12026-06-25
15Claude Opus 5max95.7%68.82026-06-25
16Grok 4.7xhigh95.7%61.62026-06-25
17DeepSeek V4 Pro95.1%53.42026-06-25
18GPT-5.6 Terramax94.9%64.42026-06-25
19Claude Opus 4.8max94.3%63.32026-06-25
20GPT-5.4xhigh94.2%64.72026-06-25
21Mistral Large 4high93.6%49.82026-06-25
22Gemini 3.7 Flashhigh93.5%63.52026-06-25
23DeepSeek V4.1 Flashmax93.3%61.22026-06-25
24GPT-5.2high93.2%56.02026-06-25
25Claude Sonnet 5xhigh92.9%59.02026-06-25
26Claude Opus 4.7xhigh92.8%57.42026-06-25
27Claude Haiku 5.5xhigh92.8%61.52026-06-25
28Grok 4.692.6%57.82026-06-25
29Gemini 3.8 Flashhigh91.6%64.42026-06-25
30Qwen3.8 Maxmax91.3%63.32026-06-25
31Muse Spark 1.2xhigh91.2%63.22026-06-25
32Gemini 3.1 Prohigh91.0%60.32026-06-25
33GPT-5.4 Nanoxhigh91.0%54.22026-06-25
34Grok 4.590.8%59.02026-06-25
35Claude Opus 4.5high90.4%59.22026-06-25
36Smaug Flash90.1%–2026-06-25
37GLM 5.289.8%51.82026-06-25
38Smaug Mini89.7%–2026-06-25
39Claude Opus 4.6high89.3%62.72026-06-25
40GPT-6 Lunamax89.1%60.02026-06-25
41GPT-5.2-Codex88.8%57.32026-06-25
42Nemotron 3 Ultra 550B A55B88.7%55.52026-06-25
43Gemini 3.5 Flashhigh88.2%63.12026-06-25
44GLM 5.387.9%–2026-06-25
45Deepseek v4 Flash Vision87.8%–2026-06-25
46GPT-5.6 Lunamax87.2%59.92026-06-25
47Muse Spark 1.1xhigh87.1%61.62026-06-25
48Claude Sonnet 4.6medium87.0%–2026-06-25
49DeepSeek V4 Flash86.8%52.22026-06-25
50Claude Haiku 5.5max86.4%60.52026-06-25
51Gemini 3.6 Flashhigh86.4%62.02026-06-25
52Qwen3.8 27B86.2%54.62026-06-25
53Qwen3.8-Flash-Next85.8%60.32026-06-25
54Qwen3.7max85.3%60.62026-06-25
55Kimi K384.4%58.82026-06-25
56Grok 4.384.3%54.62026-06-25
57Kimi K2.6thinking84.3%59.42026-06-25
58Qwen3.6 Plus83.7%58.82026-06-25
59GLM 5.3 Flash81.2%62.72026-06-25
60Qwen3.6 27B79.9%48.32026-06-25

More maths benchmarks · All LLM benchmarks

Cite as: BenchLeader, “LiveBench Mathematics leaderboard”, data as of 25 Oct 2026, https://www.benchleader.com/benchmarks/livebench_math.

LiveBench Mathematics: questions

What does LiveBench Mathematics measure?
LiveBench's mathematics questions, drawn from recent competitions and freshly generated problems, with exact answers. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads LiveBench Mathematics?
Claude Opus 5.5 leads LiveBench Mathematics with 97.1% as of 25 Oct 2026, ahead of Claude Fable 5.1 at 97.0%.
How many models have LiveBench Mathematics results?
67 model configurations have a LiveBench Mathematics result on BenchLeader, all taken from LiveBench.
Who runs LiveBench Mathematics and how often is it updated?
LiveBench Mathematics is published by LiveBench. BenchLeader re-reads the published results once a day and records the date each result was published.
Does LiveBench Mathematics count toward the BenchLeader Index?
No. LiveBench Mathematics is shown for reference but left out of the composite index.