GPT-5.6 Rumors Heat Up as Users Swear ChatGPT Suddenly Got Smarter – Decrypt

Something felt different in ChatGPT this week—and a lot of people noticed at once.

Across X, testers spent the past two days swapping screenshots and stopwatch times, all pointing to one theory: OpenAI is quietly A/B testing GPT-5.6 inside ChatGPT, swapped in for some users who select GPT-5.5 Pro.

Developer Anshu Chimala posted a side-by-side video on Thursday comparing one-shot landing pages, captioning it: “Well well well, I’m one of the lucky ones with early GPT-5.6 Pro access.”

Developer Dobroslav Radosavljevič posted on X that whatever he’s running inside Codex, OpenAI’s coding agent, “feels waaaaaaaay different than [the] 5.5 model.” Replies underneath split between believers and people calling it placebo.

The clearest pattern across the posts is time. Conor Dart, one of the many X users amplifying the rumors, ran the test with a one-prompt 3D browser game—with physics and camera controls—that took just over an hour to generate, versus GPT-5.5 Pro’s usual 10-minute mark.

“Not perfect, but for a one-prompt AI game dev test, this is seriously impressive,” Dart wrote.

AI insider Chetas Lua reported a similar slowdown testing a robotic simulation, also pretty sure his results come from OpenAI’s new model: “GPT 5.6 Pro continues to mog [Anthropic’s Fable 5] in 3D test,” he wrote. “Working on games one shot too.”

In a separate post, he noted response times stretching to 20 or 40 minutes, a pace he said hadn’t shown up since before GPT-5.5 shipped.

Not every comparison flattered the rumored model. Chris, an AI benchmarker on X, gave two models the same spaceship-building prompt—the suspected GPT-5.6 Pro worked for 87 minutes against GPT-5.5 Extra High’s 34 minutes and 42 seconds.

“As I’ve said before, based on great authority, GPT-5.6 will be an incremental/solid improvement over GPT-5.5, not a Fable killer,” he wrote, while noting Fable 5 still beat both models on the spaceship’s core geometry. “My rough expectation has been that it would trade blows with Fable 5 on some benchmarks, maybe win around half depending on the category, but not clearly surpass it overall.”

A separate post attributed to leaker Pankaj Kumar detailed leaks going further: a knowledge cutoff pushed to December 2025, a reasoning-effort setting some testers call “Juice Value” allegedly raised to 960 from 768, and SVG and 3D design generation strong enough to beat Fable 5 on some tasks.

None of it comes from OpenAI—but the details are consistent across accounts: stronger reasoning, an unfinished front end, and a release candidate nicknamed Kindle-Alpha.

An AI influencer known as Leo, citing unnamed sources, wrote in a thread that the suspected model is “now being stealth tested when 5.5 Pro is selected in ChatGPT,” at least for some Pro accounts, with a planned public launch the following Thursday, June 25.

The closest thing to an OpenAI fingerprint on any of this is a memo, not a tweet. Chief scientist Jakub Pachocki reportedly told staff the next model is a meaningful improvement over GPT-5.5, according to a report from The Information. That’s still not confirmation of A/B testing, a release date, or any spec floating around X, but it does confirm a new model has been brewing.

Decrypt reached out to OpenAI to ask whether GPT-5.6 is being tested inside ChatGPT, but the company did not respond by the time of publication.

Why OpenAI might be in a hurry

If OpenAI is rushing a new flagship model out the door, it has reasons to. China’s open-source model GLM-5.2 trails Claude Opus 4.8 by just one point on FrontierSWE—a benchmark that scores AI agents on multi-hour, open-ended engineering projects by dominance rate—while beating GPT-5.5 outright on the same test.

Anthropic, meanwhile, is dealing with damage of its own making. The company’s flagship Mythos 5 and Fable 5 models remain pulled under a U.S. export control directive issued June 12 over a disputed jailbreak vulnerability, leaving a gap at the top of the market that GLM-5.2 and a hypothetical GPT-5.6 are both positioned to fill.

If Anthropic CEO Dario Amodei and President Donald Trump make peace, then Fable 5 would be much more powerful than any other model currently available, with the quality gap between Anthropic’s top model and OpenAI’s becoming much bigger than before.

There’s also money on the table. OpenAI is reportedly weighing price cuts on the tokens it charges developers and enterprises, according to the Wall Street Journal, in anticipation of Anthropic doing the same as both companies prepare for dueling IPOs.

Whether any of it adds up to an actual GPT-5.6 release is something only OpenAI can confirm, and the company has stayed quiet through a week of leaked checkpoints and stealth-testing claims. Polymarket traders aren’t waiting around, however—contracts on a launch between June 22 and June 28 have priced as high as 89% this week.