Newsbite analysis · AI, Cloud, Programming · New Product, Software Release
The Decoder
Google launches Gemini 3.8 Live to take on OpenAI's GPT-Live-1 at a fraction of the cost
Google DeepMind has introduced Gemini 3.8 Live and Extended Thinking, two new audio models that developers can access via the Gemini API. The Gemini 3.8 Live model is revolutionary for voice agents due to its ability to support API calls in the background, handle visual inputs, and maintain conversation simultaneously. It supports over 97 languages, offering developers a competitive tool for creating voice-driven applications. While it has achieved impressive benchmarks, particularly on the Speech-to-Speech Leaderboard, Google has opted for a lower pricing strategy compared to OpenAI. This move positions Gemini as an affordable alternative, costing developers approximately $1.38 per hour of voice conversation, significantly less than the $3.00 minimum for OpenAI's equivalent services. However, the trade-off appears to be quality, as OpenAI's models are noted for more natural-sounding conversations and superior performance, especially in full duplex interactions.
ExperimentConfidence: MediumShort term horizon
Published 15 Sept 2026, 20:10 UTC1 min read
Lead image — The Decoder
Brief
NewsBite reading:Google unveils Gemini 3.8 Live to rival OpenAI's GPT-Live-1 at lower costs
TL;DR
- Google released Gemini 3.8 Live and Extended Thinking audio models.
- Models available via the Gemini API and Google AI Studio.
- Gemini 3.8 Live can handle voice and visual inputs simultaneously.
- Costs $0.005 per minute for audio input, $0.018 for output.
- Outperforms OpenAI's GPT-Live-1 on specific benchmarks.
Interesting facts
- Gemini 3.8 Live supports over 97 languages.
- The Extended Thinking model ranks first on the Speech-to-Speech Leaderboard.
- Google's pricing strategy could disrupt the competitive landscape.