Skip to content

Benchmark

Arena ELO - Polish

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

Models measured
207
Scores recorded
242
Sources
1
IQ weight
2.07

Score distribution

1,386.8
Mean
1,405.7
Median
84.1
Std dev
1,270
10th percentile
1,481.7
90th percentile
1,072
Min
1,0721,546.7
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.07 for chat models, measured over 207 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Grok 4.20Reasoning, 0309, beta· LMArena data1,452.1
52Inkling FP4· LMArena data1,451.9
53GLM 5.3 Flash· LMArena data1,451
54Grok 4.1· LMArena data1,450.5
55Qwen3.7 Plus· LMArena data1,449.3
56Qwen 3.8 27B· LMArena data1,448.7
57Qwen3.5 397B A17B· LMArena data1,448.5
58ChatGPT-4o Latest20250326· LMArena data1,447.5
59GLM 4.6· LMArena data1,447.1
60Hy3 (Deep Infra)· LMArena data1,445.4
61Kimi K2.5Thinking· LMArena data1,444.7
62Gemini 3.1 Flash Lite· LMArena data1,443.9
63GPT-5.6 TerraxHigh· LMArena data1,442.7
64Mistral Large 3· LMArena data1,441.6
65Claude Sonnet 4.6· LMArena data1,441.4
66GPT-6.1 SolMax· LMArena data1,440.9
67GPT 5.1High· LMArena data1,440.4
68GPT 5.1· LMArena data1,440.3
69Qwen3 Max Preview· LMArena data1,439.8
70Claude Opus 4.520251101· LMArena data1,438.5
71Gemini 3.5 Flash Lite· LMArena data1,437.8
72OpenAI o32025-04-16· LMArena data1,437.8
73Kimi K2.6· LMArena data1,437.3
74Claude Opus 4.5High, 32K, 20251101· LMArena data1,437.1
75GLM 4.7· LMArena data1,435.1
76DeepSeek V4 Flash· LMArena data1,435.1
77DeepSeek V3.2 ExpThinking· LMArena data1,434.9
78GLM 5· LMArena data1,434.6
79GPT-5.2 Chat20260210· LMArena data1,433.5
80Grok 3 Preview02-24· LMArena data1,432.7
81Qwen3.6 Plus· LMArena data1,432.3
82Grok 4.20 Beta 1· LMArena data1,431.7
83Grok 4.5· LMArena data1,431.5
84Gemma 4 31B IT· LMArena data1,431.4
85GPT-5 Chat· LMArena data1,431.3
86MiMo V2.6 Flash· LMArena data1,430.4
87Mistral Medium 3.12508· LMArena data1,429.5
88Mistral Medium 3.5· LMArena data1,429
89Grok 40709· LMArena data1,429
90MiMo V2 Pro· LMArena data1,428.4
91DeepSeek R10528· LMArena data1,428.1
92Claude Sonnet 5High· LMArena data1,427.9
93Grok 4 FastNon-Reasoning· LMArena data1,427.1
94Inkling Small· LMArena data1,427.1
95DeepSeek V3.1 TerminusThinking· LMArena data1,426.7
96GPT 5.2· LMArena data1,425.4
97Grok 4.6High· LMArena data1,423.4
98GPT-6 AstraMax· LMArena data1,423
99GPT 5.2High· LMArena data1,422.8
100Claude Sonnet 4.5High, 32K, 20250929· LMArena data1,422.7

Showing 51–100 of 242 results.

Data credits: LMArena, CC BY 4.0 (licence)