Skip to content

Benchmark

LiveBench paraphrase

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
1.16

Score distribution

66.7%
Mean
65.7%
Median
8%
Std dev
57.3%
10th percentile
78.1%
90th percentile
48%
Min
48%80%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.16 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 3.8 Flashhigh· LiveBench data80%
2Gemini 3.1 Pro Preview· LiveBench data79.9%
3Qwen 3.8 Flash Next· LiveBench data79.5%
4Muse Spark 1.3xhigh· LiveBench data78.6%
5Gemini 3.7 Flashhigh· LiveBench data78.1%
6Grok 4.7xhigh· LiveBench data78.1%
7Nemotron 3 Ultra· LiveBench data77.4%
8Gemini 3.5 Flashhigh· LiveBench data75.5%
9Qwen 3.8 27B· LiveBench data74.9%
10Kimi K3· LiveBench data74.4%
11Gemini 3.6 Flashhigh· LiveBench data74.3%
12Muse Spark 1.2xhigh· LiveBench data74.3%
13Qwen3.8 Max· LiveBench data73.3%
14DeepSeek V4 Flash Vision Exp· LiveBench data72.8%
15Qwen3.7 Max· LiveBench data72%
16Grok 4.6· LiveBench data71.8%
17GPT 5.5xhigh· LiveBench data71.7%
18GPT-6.1 Solxhigh· LiveBench data71%
19Inkling FP4xhigh· LiveBench data70.8%
20Muse Spark 1.1xhigh· LiveBench data69.4%
21Grok 4.5· LiveBench data69.2%
22DeepSeek V4 Pro 0813· LiveBench data68.6%
23GPT-5.2 Codex· LiveBench data67.1%
24GPT 5.4 Nanoxhigh· LiveBench data66.6%
25Claude Opus 4.7xhigh· LiveBench data65.8%
26Claude Opus 4.564k, high· LiveBench data65.7%
27Gemini 3.5 Flash Litehigh· LiveBench data65%
28Claude Opus 5.5xhigh· LiveBench data64.7%
29GLM 5.3· LiveBench data64.7%
30Grok 4.3· LiveBench data64.4%
31GPT 5.2high· LiveBench data64.1%
32Mistral Large 4high· LiveBench data64%
33GPT 5.4xhigh· LiveBench data63.4%
34Claude Opus 4.6auto, high· LiveBench data62.8%
35Claude Haiku 5.5xhigh· LiveBench data62.5%
36DeepSeek V4 Flash· LiveBench data62.4%
37Claude Sonnet 5.5xhigh· LiveBench data62.3%
38GLM 5.2· LiveBench data62.3%
39DeepSeek V4 Flash 0731· LiveBench data61.7%
40Kimi K2.6thinking· LiveBench data61.6%
41Claude Sonnet 5xhigh· LiveBench data61.3%
42Qwen3.6 Plus· LiveBench data60.6%
43Claude Sonnet 4.6auto, medium· LiveBench data59.4%
44GPT-5.4 minixhigh· LiveBench data58.9%
45Grok Build 0.1· LiveBench data58.4%
46Nemotron 3.5 Lightning 30B A3B· LiveBench data57.3%
47DeepSeek V4 Pro· LiveBench data56.2%
48MiniMax M3· LiveBench data55.9%
49Qwen3.6 27B· LiveBench data49.4%
50Kimi K2.7 Code· LiveBench data48.5%

Showing 1–50 of 51 results.