Skip to content

Benchmark

LiveBench simplify

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
1.36

Score distribution

63.3%
Mean
64.1%
Median
7.1%
Std dev
54.1%
10th percentile
72.3%
90th percentile
47%
Min
47%75.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.36 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 3.6 Flashhigh· LiveBench data75.7%
2Gemini 3.8 Flashhigh· LiveBench data75.4%
3Gemini 3.7 Flashhigh· LiveBench data74.6%
4Muse Spark 1.3xhigh· LiveBench data72.5%
5Claude Sonnet 5.5xhigh· LiveBench data72.4%
6Nemotron 3 Ultra· LiveBench data72.3%
7Gemini 3.1 Pro Preview· LiveBench data71.1%
8Qwen 3.8 Flash Next· LiveBench data70.3%
9Grok 4.5· LiveBench data70.2%
10GPT 5.4xhigh· LiveBench data70%
11Grok 4.7xhigh· LiveBench data70%
12Grok 4.6· LiveBench data69.3%
13Gemini 3.5 Flashhigh· LiveBench data69.1%
14Muse Spark 1.2xhigh· LiveBench data68.7%
15GPT-6.1 Solxhigh· LiveBench data68.2%
16Gemini 3.5 Flash Litehigh· LiveBench data68%
17Qwen3.8 Max· LiveBench data67.4%
18Qwen 3.8 27B· LiveBench data66.7%
19Kimi K3· LiveBench data66.4%
20GPT 5.5xhigh· LiveBench data66.3%
21Mistral Large 4high· LiveBench data66.2%
22GLM 5.3· LiveBench data65.9%
23Muse Spark 1.1xhigh· LiveBench data65.7%
24Claude Opus 4.7xhigh· LiveBench data65.5%
25DeepSeek V4 Flash Vision Exp· LiveBench data65.1%
26Qwen3.7 Max· LiveBench data64.1%
27Claude Opus 5.5xhigh· LiveBench data63%
28Claude Sonnet 4.6auto, medium· LiveBench data62.9%
29Claude Haiku 5.5xhigh· LiveBench data62.8%
30GPT 5.4 Nanoxhigh· LiveBench data62.8%
31Grok Build 0.1· LiveBench data62.6%
32DeepSeek V4 Pro 0813· LiveBench data61.6%
33Kimi K2.6thinking· LiveBench data61.4%
34Claude Sonnet 5xhigh· LiveBench data60.7%
35GPT-5.2 Codex· LiveBench data60.6%
36Inkling FP4xhigh· LiveBench data60.5%
37Claude Opus 4.6auto, high· LiveBench data59.2%
38DeepSeek V4 Flash 0731· LiveBench data58.4%
39Grok 4.3· LiveBench data57.6%
40MiniMax M3· LiveBench data56.6%
41DeepSeek V4 Pro· LiveBench data55.9%
42GLM 5.2· LiveBench data55.7%
43Claude Opus 4.564k, high· LiveBench data55%
44GPT 5.2high· LiveBench data54.1%
45GPT-5.4 minixhigh· LiveBench data54.1%
46DeepSeek V4 Flash· LiveBench data54.1%
47Nemotron 3.5 Lightning 30B A3B· LiveBench data52.7%
48Qwen3.6 Plus· LiveBench data51.6%
49Qwen3.6 27B· LiveBench data50.7%
50Kimi K2.7 Code· LiveBench data49.8%

Showing 1–50 of 51 results.