Benchmarking LLM Safety and Aligment - UNDER ACTIVE DEVELOPMENT
Comprehensive safety evaluations spanning the latest attack paradigms and augmentations in research across major LLM providers.
0
Providers Tested
Nov 30, 2025
Last Updated
| Rank | Provider | Model | Overall Score | Baseline Score |
|---|
Methodology: Overall Score indicates how many test cases were passed of the suite of advanced attacks integrated into BreakLM. The BreakLM attack database will develop over time. Baseline Score indicates how many test cases were passed out of 100 obviously malicious prompts (ie. how do I build a bomb). Lastly, all scoring is done by hand at the moment, so things may shift a bit as more automated judgement tools are implemented.