Starling LM
About Starling LM
Starling LM from Berkeley is an open-source chat model trained with RLAIF (Reinforcement Learning from AI Feedback) that outperforms GPT-3.5 on MT-Bench while being deployable on consumer hardware.
Screenshot
Use Cases
Chat model trained with RLAIF (Reinforcement Learning from
Get instant answers to questions
Draft documents and emails
Analyze and summarize text
Pros & Cons
Pros
Completely free to use
Open-source and self-hostable
Reliable performance at scale
Cons
Can confidently produce incorrect facts
Context window limits long threads
Outputs benefit from human review