Skip to content

Benchmark

WildBench: llama_reward.Planning & Reasoning.K=-1

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

Models measured
32
Scores recorded
32
Sources
1
IQ weight
1.67

Score distribution

26.1
Mean
36.5
Median
29.7
Std dev
-2
10th percentile
59.2
90th percentile
-61.8
Min
-61.865.5
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.67 for chat models, measured over 32 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT 4oallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data65.5
2GPT-4 Turboallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data61.2
3Claude 3.5 Sonnetallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data60.3
4Gemini 1.5 Proallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data59.2
5GPT-4-0125-previewallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data57.4
6Llama 3 70B Instructallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data52.3
7Gemini 1.5 Flashallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data50.5
8yi-largeallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data48.2
9Yi 1.5 34B Chatallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data47.9
10Claude 3 Opusallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data47.6
11Reka Coreallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data47
12Nemotron-4 340B Instructallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data46.9
13DeepSeek Coder V2 Instructallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data40.8
14Qwen2 72B Instructallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data37.2
15GLM-4-9b-chatallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data37.1
16Claude 3 Sonnetallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data36.5
17Command R Plusallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data31.9
18Claude 3 Haikuallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data24.1
19Reka Flashallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data20.6
20Llama 3 8B Instructallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data19.7
21Mistral Largeallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf, 2402· WildBench direct data18.2
22Starling LM 7B Betaallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data16.7
23Command R+allenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data12.8
24Mixtral 8x7B Instruct v0.1allenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data10.7
25Nous Hermes 2 Mixtral 8x7B DPOallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data4.8
26Tulu 2 DPO 70Ballenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data4.2
27Mistral 7B Instruct v0.2allenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data2.9
28Reka Edgeallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data1.6
29GPT-3.5 Turboallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data-2
30Llama-2-7b-chat-hfallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data-27.9
31Gemma 7B ITallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data-36.4
32Gemma 2B ITallenai/WildBench, llama_reward.Planning & Reasoning.K=-1, Llama-2-70b-chat-hf· WildBench direct data-61.8

Data credits: WildBench, Allen Institute for AI