Skip to content

Benchmark

WildBench: WB_score

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

Models measured
41
Scores recorded
41
Sources
1
IQ weight
1.53

Score distribution

41.1
Mean
44.1
Median
13.5
Std dev
26.9
10th percentile
55.8
90th percentile
-5.2
Min
-5.259.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.53 for chat models, measured over 41 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Athene 70Ballenai/WildBench, WB_score· WildBench direct data59.4
2GPT 4oallenai/WildBench, WB_score· WildBench direct data58.8
3GPT-4o miniallenai/WildBench, WB_score· WildBench direct data57.3
4GPT-4 Turboallenai/WildBench, WB_score· WildBench direct data56.1
5Mistral Large 2allenai/WildBench, WB_score· WildBench direct data55.8
6Claude 3.5 Sonnetallenai/WildBench, WB_score· WildBench direct data54.5
7Gemma 2 9B IT SimPOallenai/WildBench, WB_score· WildBench direct data54.1
8DeepSeek V2 Chatallenai/WildBench, WB_score· WildBench direct data53.8
9GPT-4-0125-previewallenai/WildBench, WB_score· WildBench direct data53.3
10Claude 3 Opusallenai/WildBench, WB_score· WildBench direct data52.1
11Llama 3 70B Instructallenai/WildBench, WB_score· WildBench direct data49.6
12yi-largeallenai/WildBench, WB_score· WildBench direct data48.9
13DeepSeek Coder V2 Instructallenai/WildBench, WB_score· WildBench direct data48.9
14Nemotron-4 340B Instructallenai/WildBench, WB_score· WildBench direct data48.8
15Gemini 1.5 Proallenai/WildBench, WB_score· WildBench direct data47.4
16Yi 1.5 34B Chatallenai/WildBench, WB_score· WildBench direct data47.4
17Mistral Nemoallenai/WildBench, WB_score· WildBench direct data46.9
18Qwen2 72B Instructallenai/WildBench, WB_score· WildBench direct data46.4
19Gemma 2 9B ITallenai/WildBench, WB_score· WildBench direct data45.4
20Claude 3 Sonnetallenai/WildBench, WB_score· WildBench direct data45.2
21Gemini 1.5 Flashallenai/WildBench, WB_score· WildBench direct data44.1
22Mistral Largeallenai/WildBench, WB_score, 2402· WildBench direct data42.3
23Command R Plusallenai/WildBench, WB_score· WildBench direct data41.6
24GLM-4-9b-chatallenai/WildBench, WB_score· WildBench direct data41.2
25Reka Coreallenai/WildBench, WB_score· WildBench direct data41
26Claude 3 Haikuallenai/WildBench, WB_score· WildBench direct data40.3
27Command R+allenai/WildBench, WB_score· WildBench direct data35.1
28Mixtral 8x7B Instruct v0.1allenai/WildBench, WB_score· WildBench direct data35
29Reka Flashallenai/WildBench, WB_score· WildBench direct data34.6
30Starling LM 7B Betaallenai/WildBench, WB_score· WildBench direct data34.2
31Nous Hermes 2 Mixtral 8x7B DPOallenai/WildBench, WB_score· WildBench direct data33.2
32Llama 3 8B Instructallenai/WildBench, WB_score· WildBench direct data33.2
33Tulu 2 DPO 70Ballenai/WildBench, WB_score· WildBench direct data32.8
34Gemma 2 2B ITallenai/WildBench, WB_score· WildBench direct data32.7
35GPT-3.5 Turboallenai/WildBench, WB_score· WildBench direct data32.3
36Mistral 7B Instruct v0.2allenai/WildBench, WB_score· WildBench direct data30.7
37Llama-2-70b-chat-hfallenai/WildBench, WB_score· WildBench direct data26.9
38Reka Edgeallenai/WildBench, WB_score· WildBench direct data23.2
39Llama-2-7b-chat-hfallenai/WildBench, WB_score· WildBench direct data15.2
40Gemma 7B ITallenai/WildBench, WB_score· WildBench direct data10.2
41Gemma 2B ITallenai/WildBench, WB_score· WildBench direct data-5.2

Data credits: WildBench, Allen Institute for AI