OpenAI has launched GPT-Live, a new voice-optimized model family that powers ChatGPT's voice mode with real-time processing capabilities unavailable in its predecessor. The models use a full-duplex architecture to interrupt users, search the web, and handle complex tasks during active conversation rather than waiting for users to finish speaking. Alongside this release, OpenAI plans to make its GPT-5.6 model series broadly available on Thursday following U.S. government approval, and its professional services division is acquiring Northslope Inc. to accelerate enterprise adoption of its AI models.
Summaries like this, in your inbox every morning.
Sign up free →What happened
OpenAI introduced GPT-Live, a family of voice-optimized AI models powering ChatGPT's voice mode. Two models launch—GPT-Live-1 (default for paid plans) and GPT-Live-1-mini (free tier)—rolling out across iOS, Android, and web with support for multiple languages. The company also plans to make GPT-5.6 broadly available on Thursday after receiving U.S. government approval.
Why it matters
GPT-Live's full-duplex architecture lets the model interrupt users, search the web, or perform tasks in real time during conversations, removing delays that plagued the previous voice system. For businesses, this means faster, more natural voice interactions. OpenAI's OpenAI Deployment Company—which raised $4 billion（約6400億円） in funding—is also acquiring Northslope Inc., a professional services firm specializing in AI applications, signaling a push to help organizations adopt these models at scale.
What to watch
GPT-Live-1 scored 75.5 on OpenAI's pleasantness evaluation, well ahead of the previous voice model. Complex prompts route to GPT-5.5, OpenAI's most advanced commercially available model. OpenAI stated that GPT-5.5 will be capable of using newer, more advanced reasoning models when they become available, with three such algorithms reportedly releasing later this week. Northslope's revenue grew sevenfold last year.
OpenAI's introduction of GPT-Live marks a shift toward real-time voice interaction, addressing a fundamental limitation of its previous system: the need to wait for users to complete their utterances before processing. The full-duplex architecture allows the model to make decisions mid-conversation—whether to interrupt, search, or delegate to a more capable model—mirroring how human conversation flows. This capability directly supports OpenAI's stated goal of making GPT-5.5 available as a backend for complex tasks routed from GPT-Live, creating a tiered system where lightweight voice queries stay on GPT-Live while sophisticated requests escalate to more powerful models.
The timing of GPT-5.6's broad release, paired with OpenAI's acquisition of Northslope Inc., suggests a coordinated push toward enterprise deployment. Northslope's specialization in building AI applications on Palantir Technologies' software, combined with its sevenfold revenue growth last year, positions OpenAI Deployment Company to serve organizations seeking to integrate these newly available models at scale. The $4 billion（約6400億円） funding raised by OpenAI Deployment Company underscores the company's commitment to capturing the professional services opportunity that accompanies major model releases. The Commerce Department's evaluation and subsequent White House approval for GPT-5.6's broader availability indicates that regulatory gatekeeping remains active for advanced AI systems, even as OpenAI moves to expand access.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
