Is GPT-6 Actually Called "Astra"? Inside OpenAI's Murky Next-Model Rumor Mill

A Second Name Enters the Chat

For most of 2026, the tech world's favorite guessing game has been simple: when does GPT-6 land, and what can it do? Lately the question has gotten messier. A cluster of AI-watcher blogs and leaker threads now claims OpenAI's next flagship won't be called GPT-6 at all — it'll reportedly be branded "Astra." Nobody at OpenAI has confirmed that, and the claim should be treated as exactly what it is: an unverified leak circulating well outside OpenAI's own channels.

That uncertainty is the real story here. The GPT-6 rumor mill has become a case study in how fast "reportedly" turns into "fact" online — and how much of what's being repeated as insider knowledge is actually aggregator sites citing other aggregator sites.

What's Actually Confirmed

Start with solid ground. OpenAI's current flagship family, GPT-5.6, launched on July 9, 2026, arriving in three tiers — Sol as the flagship, Terra as a mid-tier option, and Luna as the budget version. That rollout itself was reportedly delayed by a round of U.S. government review: multiple reports say the White House asked OpenAI to stagger the release, with the Commerce Department's Center for AI Standards and Innovation (CAISI) reportedly running additional testing before signing off on a broader launch. If accurate, that's a meaningful detail — it suggests Washington is now an active checkpoint in frontier model release timing, not just a bystander.

Sam Altman has also made a few on-record remarks worth separating from the noise. OpenAI reportedly confirmed back in October 2025 that a GPT-6-class model would not ship that year. Altman has said the gap to the next major model should be shorter than the roughly 28 months between GPT-4 and GPT-5 — though his own team, according to people who've spoken with him, rarely commits to anything more than about six months out. And in late April 2026, asked directly about GPT-6, Altman reportedly brushed off the question with a joke about shipping it "with extra goblins," offering no date, no spec sheet, nothing concrete.

Where It Gets Murky

Here's where "reportedly" needs to carry real weight. Several second-tier outlets and AI-news aggregators — not Reuters, not Bloomberg, not The Information — have been pushing a handful of competing claims about what comes after GPT-5.6:

The "Astra" name. A leak making the rounds since at least September claims OpenAI's internal codename for the next-generation model is "Astra," distinct from the GPT-6 label the public has been expecting. The outlet that surfaced it labeled it an unverified leak itself.
A 10-trillion-parameter model landing in August. This claim has already failed a basic reality check — August 2026 came and went with no such launch, and the original sourcing was thin to begin with.
A December 2026 release date. This is the one rumor OpenAI's own camp has actually pushed back on. According to reports tracking the chatter, Altman personally called the December date "fake news" when it surfaced — about as direct a denial as you'll get on any of this.

Adding to the confusion: back in March 2026, Altman confirmed a model codenamed "Spud" had finished pretraining — but that model reportedly shipped in April as GPT-5.5, not GPT-6. Several rumor roundups appear to have conflated that pretraining milestone with GPT-6 progress, which is a useful reminder of how codename soup turns into false certainty once it's been copy-pasted across enough blogs.

Why the Hedging Matters Here More Than Usual

None of this is to say GPT-6 (or Astra, or whatever it ends up being called) isn't coming. OpenAI is obviously working on something past GPT-5.6 — that much is close to certain given the company's release cadence and competitive pressure from Google, Anthropic, and xAI. What isn't certain is the name, the timeline, the parameter count, or really any of the specifics currently being reported as leaks.

This is also a useful moment to note what "sourcing" means in AI rumor coverage specifically. A lot of the GPT-6/Astra claims trace back not to named OpenAI employees, leaked internal documents, or reporters with a track record on OpenAI coverage, but to AI-content blogs summarizing each other, and to anonymous claims on social platforms that get rebroadcast as if they were confirmed. That doesn't mean they're all wrong — leaker networks around Apple and Samsung, for comparison, have real track records — but the GPT-6 rumor ecosystem right now doesn't have an equivalent to a Mark Gurman or a Ming-Chi Kuo: no single source with a verifiable hit rate that the industry has learned to trust.

The Bottom Line

If you're tracking GPT-6, the only things worth holding onto right now are: GPT-5.6 is the current OpenAI flagship, it shipped in July after a government-review delay, and nothing about the next model — not its name, not a launch window, not its scale — has been confirmed by OpenAI itself. Everything else, including the Astra rebrand, is circulating leak-blog chatter that deserves real skepticism until a reporter with an actual track record on OpenAI, or OpenAI itself, says otherwise. In a space this fast-moving, the safest read is also the most boring one: wait for the company to actually announce something.