Tomorrow is Thursday, April 23.
OpenAI has a pattern: major releases tend to land on Thursdays. This is not an internal secret but a matter of public record over the past several years.
For the next-generation model codenamed Spud, Polymarket betting data today has already assigned an 81% probability of a release, with bets concentrated on April 23.
How the Timeline Adds Up
GPT-5.5 (which may also be called GPT-6 directly; the name is not yet final) completed pretraining on March 24. This has been documented by multiple sources, and OpenAI President Greg Brockman's public interviews have corroborated it.
Based on OpenAI's pattern for several major model releases, the post-training phase (instruction fine-tuning, RLHF, safety evaluation) typically takes 4 to 6 weeks from the completion of pretraining to public release.
Counting from March 24:
- 4 weeks: April 21
- 6 weeks: May 5
Today is April 22, squarely within this window. Polymarket's reaction to the window opening has pushed the probability of a release before April 23 to 81%.
What is the market betting on? It is betting that OpenAI took the shortest path: 4 weeks of post-training, followed by a Thursday release.
What Greg Brockman Has Said
For the Spud generation, Greg Brockman's wording in interviews has been heavier than usual:
"This is not an incremental upgrade. It is a major change in how we think about model development. It contains two years' worth of research."
Sam Altman has also said that Spud could "truly accelerate the economy."
Together, these two statements say more about the expected ceiling of this generation than any benchmark screenshot. GPT-5.4 is already very capable, but if Spud were merely a "5.4 enhanced version," Brockman would not describe it as "rethinking how we develop models."
The Name Is Still Undecided
This is quite unusual — OpenAI says the official name for Spud (GPT-5.5 or GPT-6) will be determined only after internal evaluations clarify the magnitude of the performance leap relative to GPT-5.4.
If the leap feels large enough, it will be called GPT-6. If the improvement is strong but not generational, it will be called GPT-5.5.
This naming strategy itself hints at Spud's positioning: OpenAI itself is waiting for evaluation results, indicating that its performance landing point is uncertain — it could be stronger than expected, or it could be a solid iteration.
Competitive Pressure Is Real
The release window opens precisely as the impact of Anthropic's Claude Mythos 5 continues to spread.
Mythos 5 has been described as the first model to reach the trillion-parameter scale, and it is currently only open for internal testing to about 40 critical infrastructure enterprises. The emergence of such "limited-edition strong models" creates direct narrative pressure on OpenAI — using a closed high-end market to seize the title of technological leadership early.
If Spud can launch this week, it can strike at Mythos 5's weak point in terms of public availability: you have a strong model, but I have a product ready to ship.
Polymarket's 81% is not a release guarantee, but it is not a random guess either. It represents the market's collective pricing of publicly available information.
If it doesn't launch tomorrow, the window is not closed — the next reasonable node falls in the first or second week of May.
Reference sources: GPT-5.5 released on...? Trading Odds & Predictions 2026 (Polymarket); ChatGPT 5.5 aka Spud model may debut next week: Here is what to expect (Digit.in); GPT-5.5 Release Date: 70% Odds for April, CocoLoop, Spud Pretraining Done (TokenMix Blog)