Google Releases Gemini 3.8 Flash and Flash-Lite Tts Models
Something you can actually use or run today.
On September 23, 2026, Google announced the expansion of the Gemini 3.8 model family to include TTS-optimized Flash and Flash-Lite variants.
It allows developers to implement real-time conversational voice loops without experiencing jarring audio processing pauses.
This is a tactical addition to Google's fast-moving API lineup aimed at countering OpenAI's advanced voice capabilities. It provides solid utility for developer teams building consumer-facing conversational apps, though it is not an architectural revolution.
Watch for public latency comparisons checking if Flash-Lite can hold its voice quality under poor network streaming conditions.
- Provides developers with low-latency, expressive audio output models.
- Optimized specifically for real-time conversational AI interfaces.
- Distinguishes audio-focused variants from general-purpose reasoning models.
Google has officially expanded the Gemini 3.8 family with TTS-specific models.
Engineers continue to benchmark real-world latency deltas between standard mode and Extended Thinking executions.