Is ChatGPT Secretly a Genius Now? The GPT-5.6 Whispers Are Getting Impossible to Ignore

ChatGPT seemed to behave differently this week, and many users quickly picked up on it.

Across X, testers dedicated the last two days to exchanging screenshots and timing responses, all converging on a single idea: OpenAI is discreetly A/B testing GPT-5.6 within ChatGPT, making it available to some users who opt for GPT-5.5 Pro.

Developer Anshu Chimala shared a side-by-side video on Thursday, showcasing one-shot landing pages and commenting, “Well well well, I’m one of the lucky ones with early GPT-5.6 Pro access.”

Developer Dobroslav Radosavljevič also posted on X, noting that whatever he’s using within Codex, OpenAI’s coding agent, “feels waaaaaaaay different than [the] 5.5 model.” Responses varied between those convinced by the change and others dismissing it as a placebo effect.

A consistent theme across these posts is the increased processing time. Conor Dart, among the many X users fueling the speculation, conducted a test involving a single-prompt 3D browser game—complete with physics and camera controls—which took just over an hour to generate, compared to GPT-5.5 Pro’s typical 10-minute completion time.

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

AI insider Chetas Lua also noted a similar slowdown while testing a robotic simulation, expressing strong confidence that his findings are from OpenAI’s latest model: “GPT 5.6 Pro continues to outperform [Anthropic’s Fable 5] in 3D tests,” he wrote. “It’s also capable of creating games in a single attempt.”

In another post, he mentioned response times extending to 20 or 40 minutes, a pace he observed hadn’t been seen since before GPT-5.5 was released.

Not all comparisons favored the rumored model. Chris, an AI benchmarker on X, presented both models with the same spaceship-building task—the suspected GPT-5.6 Pro required 87 minutes, while GPT-5.5 Extra High completed it in 34 minutes and 42 seconds.

“As I’ve mentioned previously, based on reliable sources, GPT-5.6 will represent a steady and incremental upgrade over GPT-5.5, not a Fable killer,” he commented, adding that Fable 5 still outperformed both models concerning the spaceship’s fundamental geometry. “My general expectation has been that it would compete closely with Fable 5 on certain benchmarks, potentially winning about half depending on the specific area, but not definitively surpassing it overall.”

A separate post, attributed to leaker Pankaj Kumar, provided further alleged details: a knowledge cutoff extended to December 2025, a reasoning-effort setting, referred to by some testers as “Juice Value,” reportedly increased to 960 from 768, and SVG and 3D design generation capabilities robust enough to outperform Fable 5 on particular tasks.

None of this information originates from OpenAI directly, yet the specifics are consistent across various accounts: enhanced reasoning capabilities, an unfinished front-end, and a release candidate known by the nickname Kindle-Alpha.

An AI influencer identified as Leo, referencing unnamed sources, stated 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

The public rollout is set for Thursday, June 25.

The most concrete OpenAI signal so far isn’t a tweet—it’s an internal memo. Chief scientist Jakub Pachocki reportedly informed employees that the upcoming model represents a significant leap forward compared to GPT-5.5, according to a report from The Information. While this doesn’t amount to official confirmation of A/B testing, a release timeline, or any of the specifications circulating on X, it does verify that a new model is in the works.

Decrypt contacted OpenAI to inquire whether GPT-5.6 is undergoing testing within ChatGPT, but the company had not provided a response by the time this article went to press.

Why OpenAI may be racing to ship

If OpenAI is pushing to get a new flagship model to market quickly, there are clear motivations behind it. China’s open-source offering, GLM-5.2, trails Claude Opus 4.8 by a mere single point on FrontierSWE—a benchmark that evaluates AI agents on complex, multi-hour engineering tasks by measuring dominance rate—while outright outperforming GPT-5.5 on the same evaluation.

Anthropic, for its part, is grappling with self-inflicted setbacks. The company’s premier models, Mythos 5 and Fable 5, remain unavailable due to a U.S. export control order issued on June 12 over a contested jailbreak vulnerability. This has created a void at the high end of the market—one that both GLM-5.2 and a potential GPT-5.6 are well-positioned to occupy.

Should Anthropic CEO Dario Amodei and President Donald Trump reach an agreement, Fable 5 would emerge as substantially more capable than any model currently on offer, widening the performance gap between Anthropic’s best and OpenAI’s best considerably further.

Financial incentives are also in play. OpenAI is reportedly considering reducing the token prices it charges developers and enterprise clients, according to the Wall Street Journal, as it braces for Anthropic to do the same—both firms gearing up for competing IPOs.

Whether all of this ultimately leads to an actual GPT-5.6 launch is something only OpenAI can confirm, and the company has maintained silence throughout a week of leaked checkpoints and covert-testing rumors. Polymarket traders, however, aren’t sitting idle—contracts betting on a launch between June 22 and June 28 have climbed as high as 89% probability this week.