Skip to content

Benchmark

LiveBench theory_of_mind

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
2.85

Score distribution

76.6%
Mean
78.8%
Median
9.2%
Std dev
65.4%
10th percentile
84.6%
90th percentile
44.2%
Min
44.2%88.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.85 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
1GLM 5.3· LiveBench data88.5%
2GPT 5.4xhigh· LiveBench data88.5%
3DeepSeek V4 Flash 0731· LiveBench data86.5%
4GPT-6.1 Solxhigh· LiveBench data86.5%
5Grok 4.6· LiveBench data86.5%
6Claude Opus 5.5xhigh· LiveBench data84.6%
7DeepSeek V4 Pro 0813· LiveBench data84.6%
8GPT 5.5xhigh· LiveBench data84.6%
9Grok 4.7xhigh· LiveBench data84.6%
10Muse Spark 1.3xhigh· LiveBench data84.6%
11Claude Opus 4.6auto, high· LiveBench data82.7%
12Gemini 3.7 Flashhigh· LiveBench data82.7%
13Grok 4.5· LiveBench data82.7%
14Kimi K3· LiveBench data82.7%
15Claude Opus 4.7xhigh· LiveBench data80.8%
16Claude Sonnet 5xhigh· LiveBench data80.8%
17DeepSeek V4 Pro· LiveBench data80.8%
18Gemini 3.1 Pro Preview· LiveBench data80.8%
19Gemini 3.5 Flashhigh· LiveBench data80.8%
20GPT 5.4 Nanoxhigh· LiveBench data80.8%
21Mistral Large 4high· LiveBench data80.8%
22Muse Spark 1.2xhigh· LiveBench data80.8%
23Qwen 3.8 Flash Next· LiveBench data80.8%
24Claude Opus 4.564k, high· LiveBench data78.8%
25DeepSeek V4 Flash Vision Exp· LiveBench data78.8%
26Gemini 3.6 Flashhigh· LiveBench data78.8%
27GPT 5.2high· LiveBench data78.8%
28GPT-5.2 Codex· LiveBench data78.8%
29Qwen3.7 Max· LiveBench data78.8%
30Qwen3.8 Max· LiveBench data78.8%
31Claude Haiku 5.5xhigh· LiveBench data76.9%
32Gemini 3.8 Flashhigh· LiveBench data76.9%
33GPT-5.4 minixhigh· LiveBench data76.9%
34MiniMax M3· LiveBench data76.9%
35Muse Spark 1.1xhigh· LiveBench data76.9%
36GLM 5.2· LiveBench data75%
37Kimi K2.6thinking· LiveBench data75%
38Claude Sonnet 4.6auto, medium· LiveBench data73.1%
39DeepSeek V4 Flash· LiveBench data73.1%
40GLM 5.3 Flash· LiveBench data73.1%
41Claude Sonnet 5.5xhigh· LiveBench data69.2%
42Grok Build 0.1· LiveBench data69.2%
43Kimi K2.7 Code· LiveBench data69.2%
44Qwen3.6 Plus· LiveBench data67.3%
45Inkling FP4xhigh· LiveBench data65.4%
46Qwen3.6 27B· LiveBench data65.4%
47Grok 4.3· LiveBench data61.5%
48Nemotron 3 Ultra· LiveBench data61.5%
49Qwen 3.8 27B· LiveBench data59.6%
50Gemini 3.5 Flash Litehigh· LiveBench data50%

Showing 1–50 of 51 results.