TechOpenAI and Google's different approaches to reasoning in voice AI
The New Stack reports that OpenAI's GPT-Live-1 API uses a fast conversational model and a separate reasoning backend, cutting one customer's codebase by 80%.
TLDR
The New Stack describes two approaches to voice AI: OpenAI's GPT-Live-1 splits reasoning into a separate backend, while Google's Gemini 3.8 Live keeps reasoning inside one voice session. The outlet reports an 80% codebase reduction for one GPT-Live-1 customer and says one team deleted 23,000 lines of code.
Combined views
1.3K
2 Sources, first seen 19h ago
OpenAI and Google's different approaches to reasoning in voice AI
The New Stack reports that OpenAI's GPT-Live-1 API uses a fast conversational model and a separate reasoning backend, cutting one customer's codebase by 80%.
TLDR
The New Stack describes two approaches to voice AI: OpenAI's GPT-Live-1 splits reasoning into a separate backend, while Google's Gemini 3.8 Live keeps reasoning inside one voice session. The outlet reports an 80% codebase reduction for one GPT-Live-1 customer and says one team deleted 23,000 lines of code.