Google GenAI Python SDK v2.9.0: Model updates and API enhancements
sdkaipreviewdeprecationengineermedia
feature patch
The Google GenAI Python SDK has been updated to v2.9.0, replacing the interactions implementation with an unchanged public API surface. Key features include support for new models like gemini-3-flash-preview and Gemma 4, enhanced audio transcription capabilities, and new interaction parameters. These updates benefit developers integrating Gemini models by providing access to the latest AI capabilities and improvements.
- →Add audioOffset to VoiceActivity
- →Add gemini-3-flash-preview (gemini-3.1, gemini-3.5, gemini-4 are already mapped) to the local tokenizer map. , closes #1972
- →Add interimInputTranscription to LiveServerContent
- →Add LanguageAuto, LanguageHints, and adaptationPhrases to AudioTranscriptionConfig
- →Broaden publisher model path check to support all publishers
Features (10) ›
- Add audioOffset to VoiceActivity
- Add gemini-3-flash-preview (gemini-3.1, gemini-3.5, gemini-4 are already mapped) to the local tokenizer map. , closes #1972
- Add interimInputTranscription to LiveServerContent
- Add LanguageAuto, LanguageHints, and adaptationPhrases to AudioTranscriptionConfig
- Broaden publisher model path check to support all publishers
- Add ServiceTier to UsageMetadata
- Expose Computer Use API fields
- Gemma 4 local tokenizer support
- interaction-api Add presence_penalty, frequency_penalty, and cached_content to models.proto
- interaction-api Rename usage to total_usage in StreamMetadata
Fixes (5) ›
Add fallback for aiohttp.readline without max_line_length for backward compatibility because we still want to keep aiohttp as optional dependency , closes #2487
- Fix header
- Keep live music API keys out of websocket urls
- Make transformers an optional dependency for local tokenizers, also add other dependencies to local-tokenizer-extras
- Use .model_copy() instead of deprecated .copy()
Read the original announcement →
https://github.com/googleapis/python-genai/releases/tag/v2.9.0