GPT-Live Drops Tool Calls That Realtime Handled Fine, Developers Report After Migration
Ambitious-Pomelo-700 · reddit · 2026-09-16
A developer reports that a strict phone-agent flow (collect ID → read back → confirm → call backend tool) worked solidly on Realtime, but breaks on GPT-Live: after confirmation, the agent repeats the request or says "let me check that" and ends the turn with no tool call at all — dead air.
The author suspects the split architecture: a speech model talks while a separate delegated model decides tool calls, and the speaker can end a turn without handing off; Realtime was one model doing both. He asks whether others see this after migrating, whether any config reliably makes the delegated model fire same-turn, and whether anyone runs Live purely as a voice layer with all logic in their own backend.
More from coding & agent
- Microsoft's MemSteer turns unlabeled user histories into agent memory supervision — MicrosoftResearch · 2026-11-02
- She fired herself from her own Instagram: AI agent runs the account, views up 82.3% — heyshrutimishra · 2026-11-02
- Sourcegraph: AI writes code faster than teams can review it, driving more outages — AI Engineer · 2026-11-02
- Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate — rohanpaul_ai · 2026-11-02
- GitHub launches Agent HQ: Anthropic, OpenAI, Google coding agents in Copilot — emmanuelvivier · 2026-11-02
- DeepSeek open-sources Super-Agent, a coding agent that reads team chats to minimize human input — emmanuelvivier · 2026-11-02