Skip to content

Benchmark

LiveBench python

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
1.87

Score distribution

51.9%
Mean
50%
Median
10.7%
Std dev
40%
10th percentile
60%
90th percentile
15%
Min
15%75%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.87 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
1DeepSeek V4 Flash Vision Exp· LiveBench data75%
2Qwen 3.8 27B· LiveBench data75%
3Claude Opus 5.5xhigh· LiveBench data70%
4Qwen 3.8 Flash Next· LiveBench data70%
5Kimi K3· LiveBench data65%
6Claude Sonnet 5xhigh· LiveBench data60%
7Gemini 3.7 Flashhigh· LiveBench data60%
8GLM 5.3· LiveBench data60%
9GLM 5.3 Flash· LiveBench data60%
10GPT-5.2 Codex· LiveBench data60%
11GPT-6.1 Solxhigh· LiveBench data60%
12Grok 4.5· LiveBench data60%
13Mistral Large 4high· LiveBench data60%
14Qwen3.8 Max· LiveBench data60%
15Claude Haiku 5.5xhigh· LiveBench data55%
16GLM 5.2· LiveBench data55%
17GPT 5.2high· LiveBench data55%
18GPT 5.5xhigh· LiveBench data55%
19GPT-5.4 minixhigh· LiveBench data55%
20Grok 4.6· LiveBench data55%
21Grok 4.7xhigh· LiveBench data55%
22Kimi K2.6thinking· LiveBench data55%
23Muse Spark 1.1xhigh· LiveBench data55%
24Muse Spark 1.3xhigh· LiveBench data55%
25Claude Opus 4.6auto, high· LiveBench data50%
26Claude Sonnet 4.6auto, medium· LiveBench data50%
27DeepSeek V4 Flash 0731· LiveBench data50%
28DeepSeek V4 Pro· LiveBench data50%
29DeepSeek V4 Pro 0813· LiveBench data50%
30Gemini 3.1 Pro Preview· LiveBench data50%
31Gemini 3.5 Flashhigh· LiveBench data50%
32Gemini 3.5 Flash Litehigh· LiveBench data50%
33Gemini 3.8 Flashhigh· LiveBench data50%
34GPT 5.4xhigh· LiveBench data50%
35GPT 5.4 Nanoxhigh· LiveBench data50%
36Inkling FP4xhigh· LiveBench data50%
37Kimi K2.7 Code· LiveBench data50%
38Muse Spark 1.2xhigh· LiveBench data50%
39Claude Opus 4.7xhigh· LiveBench data45%
40Grok Build 0.1· LiveBench data45%
41Nemotron 3 Ultra· LiveBench data45%
42Qwen3.6 Plus· LiveBench data45%
43Qwen3.7 Max· LiveBench data45%
44Claude Opus 4.564k, high· LiveBench data40%
45Gemini 3.6 Flashhigh· LiveBench data40%
46Qwen3.6 27B· LiveBench data40%
47Claude Sonnet 5.5xhigh· LiveBench data35%
48DeepSeek V4 Flash· LiveBench data35%
49MiniMax M3· LiveBench data35%
50Nemotron 3.5 Lightning 30B A3B· LiveBench data35%

Showing 1–50 of 51 results.