Skip to content

Benchmark

LiveBench olympiad

Scores are percentages.

Models measured
51
Scores recorded
51
Sources
1
IQ weight
2.57

Score distribution

88.3%
Mean
89.5%
Median
5.1%
Std dev
82.8%
10th percentile
91.9%
90th percentile
58.9%
Min
58.9%92.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.57 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 4.6auto, high· LiveBench data92.2%
2Gemini 3.8 Flashhigh· LiveBench data92.2%
3Claude Opus 5.5xhigh· LiveBench data92.2%
4Gemini 3.1 Pro Preview· LiveBench data92.1%
5Gemini 3.5 Flashhigh· LiveBench data91.9%
6GPT-6.1 Solxhigh· LiveBench data91.9%
7Gemini 3.7 Flashhigh· LiveBench data91.8%
8Muse Spark 1.3xhigh· LiveBench data91.7%
9Claude Sonnet 5.5xhigh· LiveBench data91.7%
10Grok 4.7xhigh· LiveBench data91.7%
11Kimi K3· LiveBench data91.6%
12GPT 5.4xhigh· LiveBench data91.5%
13Claude Opus 4.7xhigh· LiveBench data91.4%
14GPT 5.5xhigh· LiveBench data91.4%
15DeepSeek V4 Pro 0813· LiveBench data91.3%
16Grok 4.6· LiveBench data91.2%
17Qwen 3.8 Flash Next· LiveBench data91.2%
18Qwen3.8 Max· LiveBench data91.2%
19Claude Sonnet 5xhigh· LiveBench data90.7%
20DeepSeek V4 Pro· LiveBench data90.6%
21Gemini 3.6 Flashhigh· LiveBench data90.5%
22Kimi K2.7 Code· LiveBench data90.3%
23DeepSeek V4 Flash Vision Exp· LiveBench data90.2%
24Kimi K2.6thinking· LiveBench data90%
25GLM 5.3 Flash· LiveBench data89.8%
26Claude Opus 4.564k, high· LiveBench data89.5%
27Mistral Large 4high· LiveBench data89.3%
28Grok 4.5· LiveBench data89.2%
29DeepSeek V4 Flash 0731· LiveBench data89.1%
30GLM 5.2· LiveBench data89%
31GPT 5.2high· LiveBench data88.6%
32GPT 5.4 Nanoxhigh· LiveBench data88%
33Claude Sonnet 4.6auto, medium· LiveBench data87.9%
34Muse Spark 1.2xhigh· LiveBench data87.8%
35Qwen 3.8 27B· LiveBench data87.7%
36GLM 5.3· LiveBench data87.5%
37Muse Spark 1.1xhigh· LiveBench data87.5%
38GPT-5.2 Codex· LiveBench data87%
39Claude Haiku 5.5xhigh· LiveBench data87%
40Qwen3.7 Max· LiveBench data86.9%
41Grok Build 0.1· LiveBench data86.6%
42DeepSeek V4 Flash· LiveBench data86.5%
43GPT-5.4 minixhigh· LiveBench data85.7%
44MiniMax M3· LiveBench data85.6%
45Inkling FP4xhigh· LiveBench data85.4%
46Qwen3.6 Plus· LiveBench data82.8%
47Nemotron 3 Ultra· LiveBench data82.6%
48Qwen3.6 27B· LiveBench data82.3%
49Grok 4.3· LiveBench data82.3%
50Gemini 3.5 Flash Litehigh· LiveBench data81.8%

Showing 1–50 of 51 results.