Skip to content

Benchmark

LiveBench spatial

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
1.78

Score distribution

96.3%
Mean
96%
Median
3.6%
Std dev
92%
10th percentile
100%
90th percentile
82%
Min
82%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.78 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 4.7xhigh· LiveBench data100%
2Claude Sonnet 4.6auto, medium· LiveBench data100%
3Claude Sonnet 5xhigh· LiveBench data100%
4Gemini 3.6 Flashhigh· LiveBench data100%
5Gemini 3.7 Flashhigh· LiveBench data100%
6Gemini 3.8 Flashhigh· LiveBench data100%
7Grok 4.5· LiveBench data100%
8Kimi K3· LiveBench data100%
9Muse Spark 1.1xhigh· LiveBench data100%
10Muse Spark 1.3xhigh· LiveBench data100%
11Qwen 3.8 Flash Next· LiveBench data100%
12Qwen3.6 27B· LiveBench data100%
13Qwen3.8 Max· LiveBench data100%
14Claude Opus 4.6auto, high· LiveBench data98%
15Claude Opus 5.5xhigh· LiveBench data98%
16Claude Sonnet 5.5xhigh· LiveBench data98%
17Gemini 3.1 Pro Preview· LiveBench data98%
18GPT 5.4xhigh· LiveBench data98%
19GPT 5.5xhigh· LiveBench data98%
20GPT-6.1 Solxhigh· LiveBench data98%
21Grok 4.3· LiveBench data98%
22Grok 4.6· LiveBench data98%
23Grok Build 0.1· LiveBench data98%
24Nemotron 3 Ultra· LiveBench data98%
25Qwen3.6 Plus· LiveBench data98%
26Claude Opus 4.564k, high· LiveBench data96%
27DeepSeek V4 Flash· LiveBench data96%
28DeepSeek V4 Flash Vision Exp· LiveBench data96%
29DeepSeek V4 Pro 0813· LiveBench data96%
30Gemini 3.5 Flashhigh· LiveBench data96%
31GLM 5.3· LiveBench data96%
32GLM 5.3 Flash· LiveBench data96%
33GPT 5.2high· LiveBench data96%
34Muse Spark 1.2xhigh· LiveBench data96%
35Qwen 3.8 27B· LiveBench data96%
36Qwen3.7 Max· LiveBench data96%
37Claude Haiku 5.5xhigh· LiveBench data94%
38DeepSeek V4 Pro· LiveBench data94%
39GLM 5.2· LiveBench data94%
40GPT-5.2 Codex· LiveBench data94%
41Grok 4.7xhigh· LiveBench data94%
42Inkling FP4xhigh· LiveBench data94%
43Kimi K2.6thinking· LiveBench data94%
44MiniMax M3· LiveBench data94%
45Mistral Large 4high· LiveBench data94%
46Kimi K2.7 Code· LiveBench data92%
47Nemotron 3.5 Lightning 30B A3B· LiveBench data92%
48DeepSeek V4 Flash 0731· LiveBench data90%
49Gemini 3.5 Flash Litehigh· LiveBench data90%
50GPT-5.4 minixhigh· LiveBench data88%

Showing 1–50 of 51 results.