Polymarket traders have placed more than $1 million on a single question, whether OpenAI ships GPT-5.6 between June 22 and June 28, and as of this week they’re pricing that window at 83 to 89 percent probability. The contracts crossed $960,000 in volume by June 15, according to Techtimes, and the curve has steepened since. For a company that hasn’t formally confirmed the model exists, that’s an unusually liquid prediction market.

The signals feeding it aren’t subtle. The Information, aggregated by AI Weekly, reports that OpenAI chief scientist Jakub Pachocki has told staff internally that the new model is a “meaningful improvement” over GPT-5.5, which itself was announced on April 23. A release candidate codenamed “kindle-alpha” briefly surfaced on the Design Arena testing platform before being withdrawn. And on June 19, Yahoo Tech reported ChatGPT Pro users observing longer reasoning times, stronger one-shot generation, and agentic tasks running 20 to 40 minutes, a compute footprint consistent with the period immediately preceding the GPT-5.5 launch.

Developers have noticed. “feels waaaaaaaay different than [the] 5.5 model,” wrote Dobroslav Radosavljevič, comparing recent Codex behavior to the prior baseline.

Less than two months separates GPT-5.5’s announcement from this anticipated release, a cadence that says something about how OpenAI is now organized around the IPO clock. The company filed initial paperwork with the Securities and Exchange Commission on June 8, and is reportedly weighing API pricing that would position GPT-5.6 at roughly one-third the cost of Anthropic’s Claude Fable 5. Anthropic’s flagship and its sibling Mythos were pulled on June 12 under a U.S. export-control directive, a regulatory gap OpenAI is plainly preparing to fill.

The pricing strategy and the release tempo are the same story. Greg Brockman, announcing GPT-5.5 in April, framed the trajectory as one in which models do “so much more with less guidance.” Read alongside an April 29 OpenAI post-mortem titled “Where the Goblins Came From”, which acknowledged a 175 percent rise in goblin mentions after GPT-5.1 and traced metaphor propagation across hundreds of millions of outputs, the message is that capability gains and behavioral drift are now treated as a single optimization surface.

Washington is watching the cadence too. On June 4, Representatives Jay Obernolte and Lori Trahan released a 269-page bipartisan discussion draft proposing a three-year preemption of state AI laws. The frontier-model release schedule is, in effect, setting the regulatory one.

Sources