v3.5.0

March 12, 2026 · 2 min read

Date: March 12, 2026
Version: v3.5.0
Commit: 16d85ab

Key Changes

Gemini Live integration: Added support for Google Gemini Live voice chat as a second voice model alongside OpenAI Realtime 1.5 (#389)
OpenAI Realtime 1.5 model updated: The OpenAI voice model has been updated to the latest version (gpt-realtime-1.5) (#391)

Embed assistants and lessons: Assistants and lessons can now be embedded in external websites via an <iframe>. Through the share dialog you can generate an embed code for direct integration on your own website or intranet (#390)

Google native tools for Gemini: Gemini models can now use built-in Google tools: Google Search, Code Execution, URL Context, and Google Maps. These are server-side native tools without extra function-calling overhead (#394)
File uploads for code execution: Files attached in chat are automatically uploaded to OpenAI (Code Interpreter) or Anthropic (Files API) for use during code execution (#394)
Improved agent output: Pre-tool text and post-tool text are now both correctly displayed in the chat window, resolving blank responses after tool use (#394)

macOS DMG build improved: DMG installer files for macOS can now be generated with custom license agreements (#396)