Skip to content

Benchmark

WildBench: WB_score.Creative Tasks

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

47.8
Mean
51
Median
10.7
Std dev
37.4
10th percentile
58.7
90th percentile
7.2
Min
7.260.4
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.Creative Tasks· WildBench direct data60.4
2GPT-4o miniallenai/WildBench, WB_score.Creative Tasks· WildBench direct data60.1
3GPT 4oallenai/WildBench, WB_score.Creative Tasks· WildBench direct data59.1
4Mistral Large 2allenai/WildBench, WB_score.Creative Tasks· WildBench direct data58.9
5GPT-4 Turboallenai/WildBench, WB_score.Creative Tasks· WildBench direct data58.7
6Gemma 2 9B IT SimPOallenai/WildBench, WB_score.Creative Tasks· WildBench direct data58
7GPT-4-0125-previewallenai/WildBench, WB_score.Creative Tasks· WildBench direct data57.6
8DeepSeek V2 Chatallenai/WildBench, WB_score.Creative Tasks· WildBench direct data56.4
9Claude 3.5 Sonnetallenai/WildBench, WB_score.Creative Tasks· WildBench direct data55.6
10Reka Coreallenai/WildBench, WB_score.Creative Tasks· WildBench direct data55.5
11Gemini 1.5 Proallenai/WildBench, WB_score.Creative Tasks· WildBench direct data55.1
12Mistral Nemoallenai/WildBench, WB_score.Creative Tasks· WildBench direct data54.6
13DeepSeek Coder V2 Instructallenai/WildBench, WB_score.Creative Tasks· WildBench direct data54.5
14Llama 3 70B Instructallenai/WildBench, WB_score.Creative Tasks· WildBench direct data54.3
15Yi 1.5 34B Chatallenai/WildBench, WB_score.Creative Tasks· WildBench direct data53.5
16Nemotron-4 340B Instructallenai/WildBench, WB_score.Creative Tasks· WildBench direct data53.3
17Claude 3 Opusallenai/WildBench, WB_score.Creative Tasks· WildBench direct data53
18Command R Plusallenai/WildBench, WB_score.Creative Tasks· WildBench direct data52.6
19yi-largeallenai/WildBench, WB_score.Creative Tasks· WildBench direct data51.8
20Gemini 1.5 Flashallenai/WildBench, WB_score.Creative Tasks· WildBench direct data51.7
21Gemma 2 9B ITallenai/WildBench, WB_score.Creative Tasks· WildBench direct data51
22Qwen2 72B Instructallenai/WildBench, WB_score.Creative Tasks· WildBench direct data49.9
23Mistral Largeallenai/WildBench, WB_score.Creative Tasks, 2402· WildBench direct data49.7
24GLM-4-9b-chatallenai/WildBench, WB_score.Creative Tasks· WildBench direct data47.8
25Command R+allenai/WildBench, WB_score.Creative Tasks· WildBench direct data47.4
26Claude 3 Sonnetallenai/WildBench, WB_score.Creative Tasks· WildBench direct data46.3
27Starling LM 7B Betaallenai/WildBench, WB_score.Creative Tasks· WildBench direct data43.8
28Gemma 2 2B ITallenai/WildBench, WB_score.Creative Tasks· WildBench direct data43.6
29Llama 3 8B Instructallenai/WildBench, WB_score.Creative Tasks· WildBench direct data43.6
30Claude 3 Haikuallenai/WildBench, WB_score.Creative Tasks· WildBench direct data42.9
31Mixtral 8x7B Instruct v0.1allenai/WildBench, WB_score.Creative Tasks· WildBench direct data42.8
32Tulu 2 DPO 70Ballenai/WildBench, WB_score.Creative Tasks· WildBench direct data42.7
33Reka Flashallenai/WildBench, WB_score.Creative Tasks· WildBench direct data42.4
34Mistral 7B Instruct v0.2allenai/WildBench, WB_score.Creative Tasks· WildBench direct data42.1
35Llama-2-70b-chat-hfallenai/WildBench, WB_score.Creative Tasks· WildBench direct data40
36Nous Hermes 2 Mixtral 8x7B DPOallenai/WildBench, WB_score.Creative Tasks· WildBench direct data37.9
37GPT-3.5 Turboallenai/WildBench, WB_score.Creative Tasks· WildBench direct data37.4
38Reka Edgeallenai/WildBench, WB_score.Creative Tasks· WildBench direct data36.2
39Llama-2-7b-chat-hfallenai/WildBench, WB_score.Creative Tasks· WildBench direct data29.8
40Gemma 7B ITallenai/WildBench, WB_score.Creative Tasks· WildBench direct data21.2
41Gemma 2B ITallenai/WildBench, WB_score.Creative Tasks· WildBench direct data7.2

Data credits: WildBench, Allen Institute for AI