OpenAI releases GPT-Live-1.1 voice model with full-duplex support
OpenAI launched GPT-Live-1.1 and GPT-Live-1.1 Quip on October 15, 2026, marking the next iteration of its voice AI models. The key innovation is full-duplex capability, enabling simultaneous listening and responding to mimic natural human conversation—handling pauses, interruptions, and varying speech rhythms. This replaces the previous GPT-Live-1 mini and will serve as the default voice model…
Key points
- GPT-Live-1.1 adds full-duplex voice, responding mid-conversation to match human rhythms
- Default for OpenAI Plus+ users and integrated into Codex and Dots as of October 15, 2026
- Safety tests show no regression in real-world prompts but four risk areas worsened in synthetic tests
The system card highlights a two-model architecture: GPT-Live handles voice interaction while delegating complex tasks to a backend model (e.g., GPT-5.6 Terra), inheriting its safeguards. OpenAI’s safety evaluations show no statistically significant regression in production prompts but reveal declines in four risk categories—sexual content, mental health, emotional dependency, and gore—based on synthetic tests. The company notes these are edge cases and plans further research. Alignment tests also show minor variations in adherence to restrictions and deception resistance, though all GPT-Live models outperformed the base model. OpenAI advises deploying voice AI with system-level safeguards, not just model-based ones, and emphasizes separating voice UI design from task delegation.
OpenAI's "GPT-Live-1.1" Arrives! Decoding the Capabilities and Safety Evaluations of Full-Duplex Voice AI
note.com · 16 October 2026
Loading the full article…
This text was published by note.com and written by 【AI最速情報】パープル博士. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Llama.cpp merges support for IBM Granite Switch 4.1 · 1 src
- Meta releases Muse Spark Open, an 816B-parameter open-weight multimodal model with 42B active parameters and 1M-token context · 25 src
- Reflection releases open-weight Beam 1.0 with 501B parameters and 1M-token context · 5 src
- Vercel data shows Jev hits 35% adoption in two weeks · 1 src
- Google launches Gemini 4 Argon model globally for Pro and Ultra subscribers with introductory API pricing · 33 src
Comments
via GitHub Discussions