Key Info

LiteLLM AI Gateway has added support for abliterated models from Refuseless (@refuselessai) and Abliteration (@abliteration_ai), positioned for red teaming and rigorous security testing.

Highlights

  • Two new model providers integrated: Refuseless and Abliteration
  • Abliterated models are intended for red teaming and security testing use cases
  • Available through the LiteLLM AI Gateway alongside existing providers
  • Users can route these models via LiteLLM like other supported APIs