Skip to content

Benchmark

LiveBench summarize

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
1.62

Score distribution

70%
Mean
69.3%
Median
8.6%
Std dev
58.4%
10th percentile
82.6%
90th percentile
50.4%
Min
50.4%88.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.62 for chat models, measured over 51 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini 3.1 Pro Preview· LiveBench data88.2%
2Gemini 3.8 Flashhigh· LiveBench data87.5%
3Gemini 3.7 Flashhigh· LiveBench data84.8%
4Qwen3.7 Max· LiveBench data83.8%
5Muse Spark 1.3xhigh· LiveBench data83.4%
6Gemini 3.5 Flashhigh· LiveBench data82.6%
7Muse Spark 1.2xhigh· LiveBench data80.6%
8Qwen3.8 Max· LiveBench data79.2%
9Grok 4.7xhigh· LiveBench data78.2%
10Qwen 3.8 Flash Next· LiveBench data77.4%
11Qwen 3.8 27B· LiveBench data75.9%
12Gemini 3.6 Flashhigh· LiveBench data75.9%
13GPT-6.1 Solxhigh· LiveBench data75.5%
14Nemotron 3 Ultra· LiveBench data74.9%
15Inkling FP4xhigh· LiveBench data74.8%
16GLM 5.3· LiveBench data74.5%
17Muse Spark 1.1xhigh· LiveBench data74.3%
18Grok 4.5· LiveBench data73.2%
19GPT 5.5xhigh· LiveBench data72.3%
20DeepSeek V4 Flash Vision Exp· LiveBench data72.1%
21GPT 5.4 Nanoxhigh· LiveBench data71.8%
22DeepSeek V4 Flash 0731· LiveBench data71.5%
23Claude Sonnet 5xhigh· LiveBench data71.2%
24Claude Sonnet 5.5xhigh· LiveBench data70.3%
25DeepSeek V4 Pro· LiveBench data69.5%
26Kimi K3· LiveBench data69.3%
27Claude Opus 5.5xhigh· LiveBench data69.3%
28DeepSeek V4 Flash· LiveBench data68.8%
29Gemini 3.5 Flash Litehigh· LiveBench data68.5%
30GPT 5.4xhigh· LiveBench data68.4%
31GLM 5.2· LiveBench data68.4%
32Grok 4.6· LiveBench data68.1%
33Kimi K2.6thinking· LiveBench data67.5%
34Grok Build 0.1· LiveBench data67.4%
35Claude Haiku 5.5xhigh· LiveBench data67%
36Claude Opus 4.7xhigh· LiveBench data66.3%
37DeepSeek V4 Pro 0813· LiveBench data65.7%
38Claude Opus 4.6auto, high· LiveBench data65.2%
39Qwen3.6 Plus· LiveBench data64.3%
40Claude Opus 4.564k, high· LiveBench data63.7%
41GPT-5.2 Codex· LiveBench data63.4%
42GPT 5.2high· LiveBench data63.2%
43Claude Sonnet 4.6auto, medium· LiveBench data61.6%
44Mistral Large 4high· LiveBench data61.6%
45GPT-5.4 minixhigh· LiveBench data59.6%
46Grok 4.3· LiveBench data58.4%
47Kimi K2.7 Code· LiveBench data58.3%
48GLM 5.3 Flash· LiveBench data56.5%
49MiniMax M3· LiveBench data53.9%
50Qwen3.6 27B· LiveBench data52.6%

Showing 1–50 of 51 results.