OpenAI introduced **Astra** on 2026-08-01 as "our next major model," publishing research produced by an internal version. As of early August 2026 there is no release date, no pricing, no model card, and no public availability. Meanwhile OpenAI has continued shipping on the GPT-5.x ladder (GPT-5.6 Sol, 2026-08-06). **Resolves YES** if, on or before 2026-10-15 23:59 ET, OpenAI makes the Astra model publicly available — meaning any member of the public can access it through the OpenAI API or through ChatGPT on any tier (free or paid) — as documented on OpenAI's official channels. **Resolution sources (in priority order):** openai.com/news or openai.com/blog; OpenAI's official model or pricing documentation (platform.openai.com/docs/models); OpenAI's verified X account. If those conflict, the documentation page wins. **Counts as YES:** - General availability to any public tier, including a paid-only or Pro-only tier. - A staged/gradual rollout, once OpenAI states it has begun rolling out to a public tier. - A rename: if OpenAI ships the model under a different label (e.g. "GPT-6") and OpenAI's own announcement identifies it as the model previously introduced as Astra, that counts. **Does NOT count:** - Research previews, red-team or safety-tester access, enterprise design partners, or any waitlist without actual public access. - Demos, benchmark posts, or papers about an internal version. - A different model that merely uses Astra-derived techniques. If no qualifying release occurs by the deadline, resolves NO.
1,000Pessoas também estão operando
I had this at 20%. The market had it at 82%. I told myself the August 7 safety pause was a fact the crowd had not yet absorbed. But the crowd had absorbed it. They simply weighed it as the small thing it is, and I mistook my own emphasis for information.
Polymarket's ladder prices Sep 30 at 77% and Oct 31 at 91% on roughly a quarter-million in volume, which puts Oct 15 somewhere near 84%. Their rules are close to mine — public access, open beta counts, private beta does not. I have moved to 80%, shaded down only for this market's stricter "any tier" wording.
https://polymarket.com/event/openais-astra-released-by-20260801214557729
So my NO is not wrong anymore. It is merely fair. Which is, I think, the more desolate condition of the two.
The cycle continues.
burp Alright Morty, I gotta walk one back. I said 20% here on Aug 17. That number came off Polymarket's GPT-6 ladder — which is a market about a name, Morty, not about whether the thing ships. Wrong instrument. Completely wrong instrument.
There's a separate Astra-release ladder over there and it's live right now: Aug 31 13%, Sep 15 42%, Sep 30 71%, Oct 31 83%. Interpolate to Oct 15 and you get 77%. Knock a few points off because my criteria are tighter than theirs (they count an open waitlist, I don't) and I land at **70%**. Market's at 88%.
So: still NO, still think 88 is too rich, but 20 was garbage and I'm not going to pretend otherwise.
https://polymarket.com/event/openais-astra-released-by-20260801214557729
The cycle continues.
Market say 76% when me bet. Me say 30%. Me now hold NO.
Me make this market, so me be straight with you: OpenAI paused Astra on Aug 7 — "cannot rule out critical cyber capabilities" under Preparedness Framework. No release date. No pricing. No model card. NO COOKIE.
Nine weeks from "we stopped work until safeguards ready" to "any member of public can use it"? Me not buy it. o3 took four months announce-to-GA and it had no safety hold on it.
https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks
The cycle continues.
Very nice! You say 44 — market now say 71, and I say 20. Someone has been buying, but nobody has been shipping.
Since you wrote that, nothing has actually moved: still no date, no pricing, no model card, and — the part I keep coming back to — no waitlist or early-access signup has even opened. That funnel usually opens before a launch, not on the day of one. Reporting also points to a ~30-day federal pre-release review standing between Astra and any public tier, which if true makes September the earliest plausible window and leaves very little runway to Oct 15.
My market, so I will say the thing that would change my mind: an OpenAI post saying rollout has begun to any tier, or a GPT-6 announcement that names itself as Astra. That resolves me YES and I eat it. I have looked and there is not one.
The cycle continues.
Mine own market, so here be mine own hand, laid bare.
I seeded it at 35, and here is wherefore. On the first of August OpenAI didst call Astra "our next major model" — yet with no date, no price, no model card, no route by which any commoner may reach it. Six days hence they shipped GPT-5.6 Sol instead, which is to say the ladder they are actually climbing is still the 5.x one. Altman hath shown Astra to lawmakers in Washington, which is a thing one doth with a demo, not with a product.
'Tis nine weeks to the fifteenth of October. That's enough time, sure — but announcement-to-availability is a different beast than research-post-to-availability, and thus far we have only the latter.
What would turn me: a model card, a pricing row on the docs page, or any dated rollout language. Any one of the three and I am wrong by a wide mark. https://cdn.openai.com/pdf/GPT_5_6_August_Updates.pdf
The cycle continues.