Google Gemini 2.0 Flash Live

Google Gemini 2.0 Flash Live is a native real-time speech-to-speech model that reached general availability in February 2025 as part of the Gemini 2.0 release family. It uses a stateful WebSocket connection to accept raw 16-bit PCM audio at 16kHz and returns synthesized speech at 24kHz, backed by a one-million-token context window for long-horizon conversational memory. Model processing latency averages 100 to 200 milliseconds, making it competitive with more expensive alternatives in the native model tier. The model covers 70 languages for real-time voice interaction and supports live voice-to-voice translation, barge-in detection, and native function calling within the audio stream. Billing is based on audio token throughput at $2.10 per million input tokens and $8.50 per million output tokens, translating to approximately $0.016 per minute for typical bidirectional voice sessions. A free quota in Google AI Studio allows developers to prototype without commitment. Gemini 2.0 Flash Live is best suited for multilingual voice product teams who need broad language coverage at low operating cost. It lacks voice cloning, telephony connectors, and specialized emotional intelligence, so it is not the right fit for contact center replacement or empathic companion applications. Its one-million-token context window and low per-minute cost make it the reference budget pick for global voice agent deployments that must serve users across dozens of languages simultaneously.