Skip to content

Benchmark

LiveBench zebra_puzzle

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
2.84

Score distribution

84.2%
Mean
93%
Median
18.4%
Std dev
60.4%
10th percentile
100%
90th percentile
17.3%
Min
17.3%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.84 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
1Claude Opus 5.5xhigh· LiveBench data100%
2Claude Sonnet 5.5xhigh· LiveBench data100%
3DeepSeek V4 Flash 0731· LiveBench data100%
4GPT 5.5xhigh· LiveBench data100%
5GPT-6.1 Solxhigh· LiveBench data100%
6Kimi K3· LiveBench data100%
7Qwen3.8 Max· LiveBench data100%
8DeepSeek V4 Flash Vision Exp· LiveBench data98.8%
9Qwen 3.8 Flash Next· LiveBench data98.8%
10Gemini 3.8 Flashhigh· LiveBench data98.3%
11Claude Haiku 5.5xhigh· LiveBench data98%
12Claude Opus 4.7xhigh· LiveBench data98%
13GPT 5.4xhigh· LiveBench data98%
14Muse Spark 1.3xhigh· LiveBench data98%
15GPT 5.4 Nanoxhigh· LiveBench data97.6%
16DeepSeek V4 Pro 0813· LiveBench data96.8%
17Qwen 3.8 27B· LiveBench data96.5%
18Claude Sonnet 4.6auto, medium· LiveBench data96%
19Kimi K2.7 Code· LiveBench data96%
20Muse Spark 1.1xhigh· LiveBench data96%
21Grok 4.6· LiveBench data95.5%
22Muse Spark 1.2xhigh· LiveBench data95.3%
23Gemini 3.7 Flashhigh· LiveBench data94.5%
24Claude Opus 4.6auto, high· LiveBench data94%
25Grok 4.5· LiveBench data94%
26Mistral Large 4high· LiveBench data93%
27DeepSeek V4 Pro· LiveBench data92%
28GLM 5.3· LiveBench data88.8%
29Claude Sonnet 5xhigh· LiveBench data88%
30Inkling FP4xhigh· LiveBench data88%
31Grok 4.7xhigh· LiveBench data86%
32Gemini 3.6 Flashhigh· LiveBench data85.8%
33Gemini 3.1 Pro Preview· LiveBench data85.3%
34GPT 5.2high· LiveBench data84%
35GLM 5.3 Flash· LiveBench data79.5%
36Kimi K2.6thinking· LiveBench data78.5%
37Claude Opus 4.564k, high· LiveBench data77.5%
38Gemini 3.5 Flashhigh· LiveBench data77.3%
39Qwen3.7 Max· LiveBench data74.5%
40Nemotron 3 Ultra· LiveBench data71.3%
41GPT-5.2 Codex· LiveBench data70%
42Grok Build 0.1· LiveBench data68.3%
43Qwen3.6 Plus· LiveBench data68%
44GLM 5.2· LiveBench data65.5%
45MiniMax M3· LiveBench data61%
46GPT-5.4 minixhigh· LiveBench data60.4%
47Grok 4.3· LiveBench data57.8%
48Qwen3.6 27B· LiveBench data53.8%
49DeepSeek V4 Flash· LiveBench data49.3%
50Gemini 3.5 Flash Litehigh· LiveBench data32.8%

Showing 1–50 of 51 results.