For everyday Voice users
Expect more natural overlap, better interruption handling, selectable intelligence, and visual answers. Availability can still vary by region, plan, workspace, and app version.
Latest GPT-Live updates
GPT-Live July 2026 update: what changed
GPT-Live launched in ChatGPT Voice on July 8, expanded into desktop Work and Codex workflows on July 23, and still has no generally available GPT-Live-1 API as of July 28, 2026.
Last verified:
The direct answer
The consumer product is live, the developer model is not. Free users get GPT-Live-1 mini, paid consumer plans get GPT-Live-1, and developers should continue using GPT-Realtime-2.1 while the GPT-Live-1 API remains a notification signup.
Check the API release statusJuly 2026 GPT-Live timeline
July 8 — GPT-Live launches
OpenAI introduced GPT-Live-1 and GPT-Live-1 mini for ChatGPT Voice, with full-duplex listening and speaking, interruption handling, web search, memory, text, images, and visual result cards.
July 8 — API remains pending
OpenAI said it planned to bring GPT-Live models to the API and published a notification form, but no public model ID, pricing, release date, or developer contract.
July 23 — Voice reaches Work and Codex
ChatGPT desktop added GPT-Live-powered Voice for coordinating Work and Codex tasks on macOS and Windows, with paired iOS remote access for supported accounts.
July 28 — current verified position
Live is available across supported consumer surfaces, video and screen sharing still require Advanced Voice, and GPT-Realtime-2.1 remains the documented production API path.
What is available now
Product rollout and API availability are separate contracts.
| Area | Current status | What to use |
|---|---|---|
| ChatGPT Voice | GPT-Live-1 for paid consumer plans; mini for Free | Live on supported web, iOS, Android, and desktop surfaces |
| Video and screen sharing | Not supported by Live at launch | Advanced Voice for eligible mobile subscribers |
| Work and Codex | Voice available on supported macOS and Windows desktop apps | Use Voice to start, steer, interrupt, and check tasks |
| Developer API | GPT-Live-1 is not generally available | GPT-Realtime-2.1 or GPT-Realtime-2.1 mini |
What the update means in practice
For Codex users
Voice is a coordination layer over existing agent tools and permissions, not a replacement coding model or a broader permission grant.
For developers
Do not build against an invented GPT-Live endpoint. Keep using the documented Realtime API and isolate model IDs so a future migration is deliberate.
If the update does not appear
Update ChatGPT, confirm your plan and workspace permissions, then open Settings → Voice and check whether Live is offered. Enterprise and Edu availability can require an admin to enable Voice and Early Model Access during an early-access period.
For desktop Work or Codex, confirm you are using the macOS or Windows ChatGPT app. Standalone Voice in Work and Codex is not a selectable web or mobile experience, although supported desktop sessions can use paired iOS remote access.
If you need a camera or screen-sharing input today, switch to Advanced Voice on an eligible iOS or Android account and review the GPT-Live video guide.
July update FAQ
Is GPT-Live fully released?
GPT-Live is rolling out in ChatGPT Voice, but availability still depends on plan, region, workspace settings, and app version. GPT-Live-1 is not generally available through the developer API.
What is the newest GPT-Live developer change?
There is still no public GPT-Live-1 API contract. GPT-Realtime-2.1 is the current documented model family for production voice agents.
Did GPT-Live add video in July 2026?
No. Live launched without video or screen sharing. Eligible subscribers can continue using those inputs through Advanced Voice on supported mobile apps.