Gemini 3.8 Live and 3.5 Transcribe Target Developers with Per-Minute Pricing

Google has launched two new real-time voice artificial intelligence models, Gemini 3.8 Live and Gemini 3.5 Transcribe, designed specifically for software developers building interactive conversational applications, according to reporting by ExtremeTech on September 17, 2026. The newly introduced developer-focused models are built to handle live voice interactions with low latency, expanding the company’s existing lineup of artificial intelligence tools for application builders.

Developer Pricing and Model Specifications

Under the newly announced structure, Google is charging for the use of both models on a per-minute basis, according to ExtremeTech. Gemini 3.8 Live focuses on maintaining fluid, low-latency dialogue for real-time applications, while Gemini 3.5 Transcribe specializes in speech-to-text processing for developers integrating live voice transcription features into their software. Both systems are targeted at engineering teams building voice-activated features rather than general consumer end-users.

Market Context in Real-Time Voice AI

The release places Google alongside other major technology providers offering metered voice application programming interfaces for real-time audio processing. Developers building conversational agents face strict latency requirements to ensure natural turn-taking in voice applications, a technical hurdle both Gemini 3.8 Live and Gemini 3.5 Transcribe are engineered to address. By utilizing a per-minute billing model, Google aligns its pricing with industry standards for cloud-based audio streaming and processing services.