OpenAI's next ChatGPT model, GPT-5.6, is now inside its public launch window. Polymarket prices a release between June 22 and June 28 at roughly 90 percent and a release by June 30 at about 89 percent. Pro-tier traces have appeared inside ChatGPT for some power users, and OpenAI chief scientist Jakub Pachocki has internally described the model as a meaningful improvement over GPT-5.5. The most reliable independent reporting points to a Pro variant shipping first, with the standard model and API access following.
The headline changes — based on leaked sampler outputs and developer-channel chatter, not OpenAI's own posts — are a context window pushed toward 1.5 million tokens (up from GPT-5.5's one million), an increase in the reasoning-effort budget from 768 to 960, integrated Playwright support for browser automation, and sharper long-horizon coding in Codex. None of those are confirmed by OpenAI, but together they describe a model aimed squarely at the agentic-coding workloads where Anthropic's Fable 5 currently leads SWE-bench Verified at 95.0% and where Google's still-unreleased Gemini 3.5 Pro is the other unknown.
Timing matters here. Anthropic's Fable 5 and Mythos 5 have been pulled offline globally since the June 12 export-control directive, leaving an opening in the frontier-coding tier. Gemini 3.5 Pro missed its committed June general-availability window — Sundar Pichai had targeted June at I/O on May 19 — and remains in limited Vertex preview. If GPT-5.6 ships this week, OpenAI gets a stretch of clear runway against rivals that are either offline or delayed, just as enterprise procurement cycles begin Q3 planning.
Takeaway for learners: model release cadence has compressed to the point where prediction markets are more accurate than vendor PR calendars. If you are building on top of these APIs, do not architect around the current frontier — architect around the model two releases ahead. The version of GPT, Claude, or Gemini you call today will be unrecognisable in six months, and the engineers who win are the ones who build interfaces and evaluation harnesses that survive that churn.