Key Info

LiteLLM Gateway now offers live speech translation between languages through gpt-realtime-translate, available from OpenAI and Azure, with sessions billed per second of audio sent.

Highlights

  • Live speech-to-speech translation is exposed through the LiteLLM Gateway.
  • The feature uses gpt-realtime-translate from OpenAI and Azure.
  • Billing is based on seconds of audio sent per session.
  • Recent merges include Claude prompt caching for OpenAI-built apps, Claude PDF reading from web links, retired-model replacement display, and a 30-second cap on slow spend-log lookups in the admin dashboard.