Will OpenAI release its next major GPT model before December 2026?
Resolves YES if a new flagship GPT-series model becomes generally available to the public before December 1, 2026. Point releases, distilled variants, and research previews without general availability do not count.
Set a lens and this page will tell you why the question matters for you, plus a second reading if you weight one agent more. Set your lens
Probability over time
Running the forecast
- Gathering evidence…
- Checking historical base rates…
- Reading recent developments…
- Challenging assumptions…
- Comparing against market consensus…
- Auditing sources for circular reporting…
- Aggregating agent forecasts…
Why we disagree
Contrary sits 18 points below consensus. The market is pricing an imminent flagship, but this question requires general availability before December 1, and the most reliable leading indicator, a public developer preview, is absent after 90 days. Three bullish reports share a single original source, so the evidence base is narrower than its headline count.
- The question turns on general availability, but most reporting describes announcements, the two have diverged by months before.
- What counts as a 'major' model is contested; a strongly-marketed point release could trigger a resolution dispute.
- A late-November launch window is plausible and sits within days of the cutoff.
What would change the forecast?
Toggle a scenario to see how Contrary would revise its probability if that development were confirmed.
Time Machine
Rewind to any point and see only what was knowable then. Nothing published after the selected date is used.
The widest disagreement of the year. Contrary is 18 points below consensus, driven by a missing preview and a narrow source base.
- Developer changelog confirms no preview in 90 days
- Skeptic audit collapses three reports into one original source
Circular reporting identified; bullish evidence re-weighted downward.
Sources
Every factor traces to a source. The Skeptic Agent reduces the influence of weak, unverified, or duplicated reporting.
14 new post-training evaluation and safety roles posted
Two rival labs ship frontier updates within three weeks
Sources say next flagship model slips toward early 2027
Skeptic Agent: Single anonymous source, no corroborating reporting
Screenshot claims an internal model card leaked
Skeptic Agent: Image unverifiable; original post deleted within hours
What we know so far about OpenAI's next model
Skeptic Agent: Restates The Information reporting; adds no new sourcing