Skip to content

Benchmark

LiveBench

Math benchmarks measure how well models solve maths problems. Scores are percentages.

Models measured
71
Scores recorded
87
Sources
2
IQ weight
2.39

Score distribution

72.4%
Mean
74.3%
Median
9.1%
Std dev
66.1%
10th percentile
79.5%
90th percentile
29.6%
Min
29.6%84.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.39 for chat models, measured over 70 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1o3 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.6%
2Muse Spark 1.3overall, xhigh· LiveBench data82.4%
3Claude Opus 5.5overall, xhigh· LiveBench data82.1%
4GPT-6.1 Soloverall, xhigh· LiveBench data81.8%
5GPT 5.5overall, xhigh· LiveBench data80.8%
6GPT 5.5· LLM Stats data80.7%
7GPT 5.4· LLM Stats data80.3%
8Gemini 3.1 Pro Preview· LLM Stats data79.9%
9Gemini 3.7 Flashoverall, high· LiveBench data79.9%
10Qwen3.8 Maxoverall· LiveBench data79.5%
11Kimi K3overall· LiveBench data79.5%
12Grok 4.6overall· LiveBench data79%
13Muse Spark 1.2overall, xhigh· LiveBench data78.9%
14GPT 5.4overall, xhigh· LiveBench data78.8%
15Grok 4.7overall, xhigh· LiveBench data78.3%
16Claude Fable 5· LLM Stats data78.3%
17Claude Sonnet 5.5overall, xhigh· LiveBench data78.2%
18DeepSeek V4 Pro 0813overall· LiveBench data78.2%
19Gemini 3.1 Pro Previewoverall· LiveBench data78%
20DeepSeek V4 Flash Vision Expoverall· LiveBench data77.7%
21Gemini 3.8 Flashoverall, high· LiveBench data77.5%
22Qwen 3.8 Flash Nextoverall· LiveBench data77.3%
23Claude Opus 4.8· LLM Stats data77.2%
24Qwen 3 235B A22B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim77.1%
25Grok 4.5overall· LiveBench data77.1%
26Claude Opus 4.7overall, xhigh· LiveBench data77%
27Claude Opus 4.7· LLM Stats data76.9%
28GLM 5.3overall· LiveBench data76.6%
29Claude Sonnet 5overall, xhigh· LiveBench data76.6%
30Kimi K2 Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim76.4%
31Claude Opus 4.6· LLM Stats data76.3%
32Muse Spark 1.1overall, xhigh· LiveBench data76%
33Claude Opus 4.5· LLM Stats data76%
34Qwen 3.8 27Boverall· LiveBench data75.8%
35Claude Sonnet 4.6· LLM Stats data75.5%
36Gemini 3.5 Flashoverall, high· LiveBench data75.4%
37GPT 5.2overall, high· LiveBench data75.2%
38Claude Opus 4.6overall, auto, high· LiveBench data75.1%
39Gemini 3.5 Flash· LLM Stats data75%
40Qwen3 32B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim74.9%
41DeepSeek V4 Flash 0731overall· LiveBench data74.8%
42GPT 5.2· LLM Stats data74.8%
43Gemini 3.6 Flashoverall, high· LiveBench data74.5%
44GPT-5.2 Codex· LLM Stats data74.3%
45Qwen3 30B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim74.3%
46Qwen3.7 Max· LLM Stats data74.3%
47Qwen3.7 Maxoverall· LiveBench data74.1%
48GPT-5.2 Codexoverall· LiveBench data74%
49Claude Sonnet 4.6overall, auto, medium· LiveBench data73.4%
50Gemini 3 Pro Preview· LLM Stats data73.4%

Showing 1–50 of 87 results.

Data credits: Data by LLM Stats