LMArena · Agent

LMArena · Agent is run by LMArena. It has ranked 53 models, of which the catalogue holds 33, scoring from 1 to 93 on Net Improvement (%).

measured by LMArena · the board itself

Place	Model	Metric	Score
2ndof 53	Claude Opus 5.5	Net Improvement (%)	15.53
4thof 53	GPT 6 Astra	Net Improvement (%)	13.09
5thof 53	Claude Fable 5.1	Net Improvement (%)	12.66
7thof 53	Claude Sonnet 5.5	Net Improvement (%)	11.95
8thof 53	GPT 6.1 Sol	Net Improvement (%)	11.72
11thof 53	GPT 6 Sol	Net Improvement (%)	9.92
13thof 53	Claude Fable 5	Net Improvement (%)	8.94
14thof 53	Claude Opus 5	Net Improvement (%)	8.12
16thof 53	Gemini 3.6 Flash	Net Improvement (%)	7.79
17thof 53	Gemini 3.1 Pro Preview	Net Improvement (%)	7.79
18thof 53	MiMo-V2.5	Net Improvement (%)	7.78
20thof 53	Claude Opus 4.8	Net Improvement (%)	7.13
22ndof 53	Mistral Large 4	Net Improvement (%)	6.56
23rdof 53	GPT-5.6 Sol	Net Improvement (%)	6.23
26thof 53	Muse Spark 1.2	Net Improvement (%)	5
27thof 53	Claude Sonnet 5	Net Improvement (%)	4.91
28thof 53	GPT-5.5	Net Improvement (%)	4.38
29thof 53	Muse Spark 1.3	Net Improvement (%)	4.04
30thof 53	DeepSeek V4.1 Flash	Net Improvement (%)	3.8
36thof 53	Grok 4.7	Net Improvement (%)	2.83
37thof 53	Gemini 3.8 Flash	Net Improvement (%)	2.75
38thof 53	qwen3.8	Net Improvement (%)	2.35
39thof 53	GLM-5.3	Net Improvement (%)	2.15
40thof 53	Gemini 3.7 Flash	Net Improvement (%)	2.1
42ndof 53	GPT 6 Luna	Net Improvement (%)	1.93
43rdof 53	GPT-5.6 Luna	Net Improvement (%)	1.48
44thof 53	DeepSeek V4 Pro	Net Improvement (%) (high)	1.34
46thof 53	GPT-5.4	Net Improvement (%)	1.19
47thof 53	Grok 4.6	Net Improvement (%)	1.05
49thof 53	openai-gpt	Net Improvement (%)	93
50thof 53	GPT-5.6 Terra	Net Improvement (%)	88
51stof 53	Qwen3.8 Flash	Net Improvement (%)	72
53rdof 53	Step 5	Net Improvement (%)	1

Place

Model

Metric

Score

2ndof 53

Net Improvement (%)

15.53

4thof 53

Net Improvement (%)

13.09