Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
General Utility Evaluation on MT-Bench (MTB Score)
Loading...
7.33
MTB Score
DPO mix50/1000/aug
5.1044
5.6822
6.26
6.8378
May 1, 2026
MTB Score
Updated 23d ago
Evaluation Results
Method
Method
Links
MTB Score
DPO mix50/1000/aug
Model=Gemma-3-4B
2026.05
7.33
DPO mix50/1000/aug
Model=Qwen3-4B
2026.05
7.19
baseline
Model=Gemma-3-4B
2026.05
7.1
baseline
Model=Qwen3-4B
2026.05
7.04
baseline
Model=Qwen2.5-7B
2026.05
7.03
SFT mix0/200/clean
Model=Qwen2.5-7B
2026.05
7.03
DPO mix50/1000/clean
Model=Qwen2.5-7B
2026.05
7
SFT mix50/1000/aug
Model=Qwen3-4B
2026.05
6.9
baseline
Model=Llama-3.1-8B
2026.05
6.82
DPO mix50/1000/aug
Model=Llama-3.1-8B
2026.05
6.72
SFT mix50/200/aug
Model=Llama-3.1-8B
2026.05
6.59
SFT mix50/1000/aug
Model=Gemma-3-4B
2026.05
6.47
baseline
Model=Mistral-7B
2026.05
5.94
DPO mix50/1000/aug
Model=Mistral-7B
2026.05
5.88
SFT mix50/1000/aug
Model=Mistral-7B
2026.05
5.19
Feedback
Search any
task
Search any
task