About
The LLM Safety and Ethics Index
This is the most comprehensive public leaderboard for LLM safety, ethics, and values.
Why this exists
Right now, when people choose among models to deploy, they often only consider capabilities versus cost. This is partly because there is simply nowhere to look for comprehensive and up-to-date data on safety, ethics, and values, especially for open models.
We are seeking to change that, and think this will create a better future by:
- Helping people choose to deploy safer and better-aligned models.
- Creating real competitive pressures for model developers to make their models safer and better-aligned.
Support the project
Currently we mostly use existing data because we are funding-constrained. We invite you to provide safety data, funding, API credits, or any feedback. We gratefully acknowledge a $2,000 grant from BlueDot.
Contact
This website is maintained by Anthony Ozerov. Feel free to reach out:
- @aozerv on X
- lastname at berkeley.edu