Google Launches Gemini 3.1 Flash Live — Its "Highest-Quality Audio Model Yet"
Google today launched Gemini 3.1 Flash Live, a new voice model the company describes as the "biggest upgrade yet" to its Gemini Live platform. The model brings meaningfully improved precision and dramatically lower latency — with response times under 200 milliseconds — making AI-driven voice conversations feel noticeably more natural and fluid. Tonal recognition, conversational flow, and overall audio fidelity have all seen significant upgrades in this release.
The new model now powers both Gemini Live and Search Live, the latter of which Google simultaneously expanded globally. Developers can access Gemini 3.1 Flash Live today in preview through the Live API endpoint, making it immediately usable for applications that depend on low-latency, real-time spoken interaction. The release underscores a clear strategic bet from Google: voice is the next frontier for AI assistants, and the company is investing heavily to lead it.
Real-time voice has become one of the most intensely competitive arenas in AI, with OpenAI's Advanced Voice Mode and other players racing to close the gap. Google's sub-200ms latency benchmark, paired with improvements in tonal awareness, positions Gemini 3.1 Flash Live as a genuine step forward — not just an incremental update. For developers building voice-first products on the Gemini ecosystem, today's preview release opens a new level of capability.