Starling-7B
Open in the interactive archive →
Download the covers: Floppy card · VHS cover · Vinyl cover
A to Z
- A · App
- No
- B · Built By
- UC Berkeley
- C · CEO & Politics
- N/A (academic research) — No public partisan activity documented
- D · Date
- 2023
- E · Expertise
- Intermediate+
- F · Free Tier
- Yes
- G · Generates
- Text only
- H · Deployment
- Cloud + Local (open weights)
- I · Inputs
- Text
- J · Jurisdiction
- USA
- K · Key Staffers
- Banghua Zhu (Co-lead author), Evan Frick (Co-lead author), Tianhao Wu (Co-lead author), Jiantao Jiao (Senior author)
- L · Languages
- English-primary
- M · Model Type
- LLM
- N · Safety Issues
- None reported
- O · Open Source
- Yes — open weights
- P · Pricing
- Free (open weights, Apache 2.0 license on condition the model is not used to compete with OpenAI)
- Q · AIDb Rating
- [UNKNOWN]
- R · What Makes It Special
- Starling-7B, an open language model from UC Berkeley researchers trained with reinforcement learning from AI feedback: Starling-LM-7B-alpha fine-tunes Openchat 3.5, itself based on Mistral-7B-v0.1, using the new Starling-RM-7B-alpha reward model, trained on Berkeley's GPT-4-labeled Nectar ranking dataset, and advantage-induced policy alignment. It scored 8.09 on MT-Bench with GPT-4 as judge, behind only GPT-4 and GPT-4 Turbo at release; its Apache 2.0 license bars use to compete with OpenAI.
- S · Stock / Status
- University
- T · Training Cutoff
- [UNKNOWN]
- U · Users
- [UNKNOWN]
- V · Voice Chat
- No
- W · Weights
- Downloadable
- X · X Handle
- [UNKNOWN]
- Y · Year Founded
- [UNKNOWN]
- Z · Bottom Line
- Open-weights LLM from UC Berkeley, released 2023.
Data completeness 81% · Last verified 2026-05-12 · Primary source ↗
Cite this
“Starling-7B.” AIDb — The AI Database. Verified 2026-05-12. https://theaidatabase.com/m/starling-7b.html