Skip to content

Benchmark

Arena ELO - Industry Software and It Services

People compare model answers and choose which they prefer. Higher scores are better.

Models measured
366
Scores recorded
423
Sources
1
IQ weight
1.29

Score distribution

1,322.6
Mean
1,374.8
Median
161.3
Std dev
1,081.5
10th percentile
1,486.1
90th percentile
719.3
Min
719.31,552.3
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.29 for chat models, measured over 366 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
101GLM 4.7· LMArena data1,453.4
102DeepSeek V4 Flash· LMArena data1,452.6
103Grok 4.20 Beta 1· LMArena data1,452.6
104Claude Opus 4.120250805· LMArena data1,451.4
105Mistral Large 3· LMArena data1,451.2
106Grok 4.1Thinking· LMArena data1,450.8
107Gemini 3.5 Flash Lite· LMArena data1,450.5
108Grok 4.1· LMArena data1,449.7
109DeepSeek V3.2Thinking· LMArena data1,449.6
110Inkling Small· LMArena data1,447.8
111nova-experimental-chat12-10· LMArena data1,447.4
112Qwen3 Next 80B A3B Instruct· LMArena data1,447
113ERNIE-5.0-Preview1203· LMArena data1,446.8
114Qwen3 235B A22B Instruct 25072507· LMArena data1,446.7
115Kimi K2 Thinking· LMArena data1,446.5
116MiniMax M2.7· LMArena data1,445.4
117Gemini 3 FlashMinimal· LMArena data1,444.7
118DeepSeek V3.2· LMArena data1,443.9
119Inkling 2 Small· LMArena data1,443.7
120DeepSeek-V4-Flash PreviewHigh· LMArena data1,443.1
121MiMo V2 FlashNon-Thinking· LMArena data1,443.1
122GPT-6.2 AstraMax· LMArena data1,443
123Qwen3 VL 235B· LMArena data1,442.2
124nova-experimental-chat11-10· LMArena data1,441.5
125Claude Haiku 5.5High· LMArena data1,441
126GLM 4.5· LMArena data1,440.5
127NVIDIA Nemotron 3.5 Super VL 120BNVFP4· LMArena data1,440.4
128nova-experimental-chat26-01-10· LMArena data1,439.9
129Claude Haiku 4.520251001· LMArena data1,439
130DeepSeek V3.2 ExpThinking· LMArena data1,438.2
131Grok 3 Preview02-24· LMArena data1,437.5
132GPT 5.1· LMArena data1,436.8
133Qwen3.5 122B-A10B· LMArena data1,436.6
134Mistral Medium 3.12508· LMArena data1,436.5
135DeepSeek V3.2 Exp· LMArena data1,436.2
136GPT-5.4 miniHigh· LMArena data1,436.2
137Qwen3 Max2025-09-23· LMArena data1,435.7
138ERNIE-5.0-Preview1022· LMArena data1,434.7
139DeepSeek R10528· LMArena data1,434.6
140Step 3.5 Flash· LMArena data1,434.5
141GPT 5.2High· LMArena data1,434.2
142GPT-6.2 SolMax· LMArena data1,433.7
143GPT-6 SolMax· LMArena data1,432.9
144GPT 5.2· LMArena data1,432.4
145GPT-5.5 Instant· LMArena data1,432.2
146Qwen3.5 27B· LMArena data1,431.5
147Qwen 3 235B A22B2507, thinking· LMArena data1,431.1
148Grok 4 FastNon-Reasoning· LMArena data1,430.7
149ChatGPT-4o Latest20250326· LMArena data1,428.6
150Solar Pro 4· LMArena data1,427.7

Showing 101–150 of 423 results.

Data credits: LMArena, CC BY 4.0 (licence)