ASTRA LAUNCH WATCH · SEPTEMBER 3
Has Astra launched? Here is the evidence.
Released September 3, 2026. Access is rolling out in stages.
This is the timestamped answer for public access, GPT-6 and GPT-5.7 naming, ChatGPT/API availability, and the live market signal. Contract definitions, access criteria, timestamps, and rumor classifications stay visible so a market move or technical report cannot quietly become a fact.
Current answer
OpenAI has released GPT-6 Astra. The market cards below now preserve launch-window history and explicit resolution rules; they remain different questions from the official access record.
Publication rule: the release is confirmed by authoritative OpenAI surfaces; rumors, market movement, screenshots, and API-routing behavior still do not establish details beyond the published record.
THE DISTRIBUTION PRODUCT
One release question. Six evidence lanes.
Official signals, credible reporting, market movement, technical signals, classified rumors, and editorial interpretation stay in separate lanes. Each entry says what changed, what it does not prove, and how it affects the estimated release window.
GPT-6 Astra is released; access routes expand
- What changed
- OpenAI publishes the launch safety overview, model documentation, and system card. The official model record names gpt-6-astra, publishes API pricing and limits, and should now be read alongside the route-specific ChatGPT, Work, Codex, API, Copilot, and Foundry access records.
- What it does not prove
- It does not mean every ChatGPT plan, API account, region, or tool surface is available immediately. It also does not replace independent testing of real workflows.
- Release-window effect
- Closes the release-date question and opens the access-verification window: link the official docs, test eligible accounts, and report staged availability accurately.
Astra moves from “may meet” to OpenAI’s Critical cyber threshold
- What changed
- OpenAI now says Astra meets its Critical cybersecurity capability threshold, reports 100% on ExploitBench and two zero-days during evaluation, and says Astra will be available soon with a system card at launch.
- What it does not prove
- It does not publish an exact date, public ChatGPT or API access, pricing, quotas, or default-production performance. The results reflect Daybreak Blue access and OpenAI’s evaluation conditions.
- Release-window effect
- Raises launch-readiness urgency and strengthens the imminence signal, but leaves the calendar uncommitted.
Sources: OpenAI ↗
Controlled defender access is part of the Astra cyber path
- What changed
- OpenAI describes Daybreak Blue as access for approved defenders while the cyber-defense window narrows, giving the safety story a concrete controlled-access surface.
- What it does not prove
- Daybreak Blue is not a public Astra waitlist, a general ChatGPT launch, or an API access announcement. Approved-defender access cannot be generalized to ordinary readers.
- Release-window effect
- Supports a staged-access interpretation: prepare for constrained rollout and verification rather than assume everyone gets access at once.
Sam Altman is reported to describe Astra as near human parity at computer use
- What changed
- Sources News quotes Sam Altman saying Astra feels like it has reached roughly human parity at using computers, creating a high-intent computer-use question beyond the mathematics announcement.
- What it does not prove
- A reported interview statement is not an independent task battery, a product specification, a safety guarantee, or proof that readers can use Astra today.
- Release-window effect
- Increases demand for practical computer-use documentation and a bounded test plan; it supplies no release date.
Sources: Sources News ↗ · Reuters ↗
Astra-specific markets lean toward an imminent window
- What changed
- At the latest capture, Polymarket showed 84.5% for release by September 3 and 91% by September 4 using CLOB midpoints. Kalshi’s separate Astra contract showed 97.5% before September 18.
- What it does not prove
- These are separate market definitions, not an average and not an OpenAI probability. A quote cannot prove public access, a release commitment, or that a GPT-6-named model is Astra.
- Release-window effect
- Supports an imminent-window operating posture while preserving the official “soon, no date” answer and the public-access gate.
Sources: Polymarket ↗ · Kalshi ↗ · Polymarket ↗ · Kalshi ↗
API-routing reports are a lead, not a public Astra release
- What changed
- A third-party technical report claims to see gpt-6-astra routing behavior. It is useful to monitor because it speaks to implementation signals, but it is not an official product or access record.
- What it does not prove
- It does not establish a public model identifier, working access, ChatGPT availability, pricing, or that the observed routing is stable production behavior. The official model catalog and ChatGPT release notes remain the publication gate.
- Release-window effect
- Keep the prelaunch gate active; if the signal changes, re-test through an official product or API surface before changing the site mode.
Sources: Net.Coffee ↗ · OpenAI ↗ · OpenAI ↗
Specific launch-day rumors remain unsubstantiated
- What changed
- The official “available soon” language increases attention on date claims, but no specific day is promoted to the verified record. First-party intent still describes direction, not a calendar commitment.
- What it does not prove
- It does not prove that a date rumor is true, that a market quote is correct, or that an OpenAI waitlist exists.
- Release-window effect
- No change to the publishable date window; keep exact-day claims out of headlines until an official access signal exists.
Sources: Sam Altman / X ↗ · OpenAI ↗
Run launch operations as if the window is active, without publishing a date
- What changed
- The combination of official “soon” language, a promised launch system card, controlled cyber access, and fresh practical demand justifies completing the email, source, benchmark, and distribution surfaces now.
- What it does not prove
- This is an operating assumption, not an OpenAI statement, a probability, or permission to switch the site into launch mode.
- Release-window effect
- Moves work from speculative planning to launch readiness: verify access first, then update facts, send the approved alert, and begin the frozen battery.
MARKET SIGNAL, NOT FACT
What prediction markets are pricing
Prices show what traders are willing to pay for a contract under that venue’s rules. They are not an OpenAI forecast, a confirmed schedule, or a substitute for primary-source evidence.
OpenAI's Astra released by...?
Astra, or a model confirmed to be the same model, publicly accessible by the listed date.
Resolution: Open beta or an open rolling waitlist can qualify; closed beta or private access cannot. GPT-5.7 or GPT-6 only count if confirmed to be the Astra model.
Read this carefully: These are market-implied prices, not OpenAI probabilities or a release commitment. The “same model” and credible-reporting clauses still require judgment at resolution.
Data quality: public market observation.
Observed basis: CLOB midpoint. Percentages are shown for the Yes outcome; they are not forecasts.
Open the Polymarket sourceWhen will OpenAI release Astra?
Astra-specific Kalshi contracts asking whether OpenAI releases Astra before each listed date.
Resolution: Each contract resolves from the filed Astra market terms and Kalshi’s public-release criteria; a generic GPT-6 release is not automatically Astra.
Read this carefully: These are market-implied prices, not OpenAI probabilities or commitments. The contract terms and public-access definition control resolution.
Data quality: public market observation.
Observed basis: Yes bid/ask midpoint. Percentages are shown for the Yes outcome; they are not forecasts.
Open the Kalshi sourceWhen will OpenAI release GPT-6?
A related GPT-6 release series. It is not an Astra contract and does not resolve the same question.
Resolution: The filed GPT-6 terms require a public OpenAI model whose name starts with “GPT-6.” A GPT-5.7 release would not qualify just because it might be Astra.
Read this carefully: Use this as a related naming signal only. Do not average it with Polymarket or describe it as the probability of an Astra release.
Data quality: public market observation.
Observed basis: Yes bid/ask midpoint. Percentages are shown for the Yes outcome; they are not forecasts.
Open the Kalshi sourcePRESERVED OBSERVATIONS
History starts here
The tracker currently contains 37 preserved observations. Future refreshes append API data rather than overwrite the record.
| Observed UTC | Venue | Outcome | Shown probability | Volume |
|---|---|---|---|---|
| 2026-08-18T17:02:00Z | Polymarket | August 31 | 21% | $231,261 |
| 2026-08-18T17:02:00Z | Polymarket | September 15 | 54% | $231,261 |
| 2026-08-18T17:02:00Z | Polymarket | September 30 | 72% | $231,261 |
| 2026-08-18T17:02:00Z | Polymarket | October 31 | 84% | $231,261 |
| 2026-08-18T17:02:00Z | Kalshi | Before September 1 | Not quoted | $401,037 |
| 2026-08-18T17:02:00Z | Kalshi | Before October 1 | 28% | $401,037 |
| 2026-08-18T17:02:00Z | Kalshi | Before November 1 | 48% | $401,037 |
| 2026-08-18T17:02:00Z | Kalshi | Before 2027 | 71% | $401,037 |
| 2026-08-19T23:17:00Z | Polymarket | August 31 | 12% | $270,283 |
| 2026-08-19T23:17:00Z | Polymarket | September 15 | 38% | $270,283 |
| 2026-08-19T23:17:00Z | Polymarket | September 30 | 73% | $270,283 |
| 2026-08-19T23:17:00Z | Polymarket | October 31 | 87% | $270,283 |
| 2026-08-19T23:17:00Z | Kalshi | Before September 1 | Not quoted | $405,449 |
| 2026-08-19T23:17:00Z | Kalshi | Before October 1 | 25% | $405,449 |
| 2026-08-19T23:17:00Z | Kalshi | Before November 1 | 45% | $405,449 |
| 2026-08-19T23:17:00Z | Kalshi | Before 2027 | 75% | $405,449 |
| 2026-08-30T23:12:00Z | Polymarket | August 31 | 2% | $156,130.96 |
| 2026-08-30T23:12:00Z | Polymarket | September 15 | 72.5% | $65,386.25 |
| 2026-08-30T23:12:00Z | Polymarket | September 30 | 85% | $88,073.27 |
| 2026-08-30T23:12:00Z | Polymarket | October 31 | 93% | $36,256.46 |
| 2026-09-03T11:25:00Z | Polymarket | September 3 | 84.5% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 4 | 91% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 9 | 93.75% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 11 | 96.15% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 15 | 97.35% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 25 | 98.2% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | September 30 | 98.75% | $425,470.676 |
| 2026-09-03T11:25:00Z | Polymarket | October 31 | 98.85% | $425,470.676 |
| 2026-09-03T11:25:00Z | Kalshi | September 18 | 97.5% | $169,301.53 |
| 2026-09-03T11:25:00Z | Kalshi | October 1 | 98.5% | $169,301.53 |
| 2026-09-03T11:25:00Z | Kalshi | October 16 | 98.5% | $169,301.53 |
| 2026-09-03T11:25:00Z | Kalshi | November 1 | 99.5% | $169,301.53 |
| 2026-09-03T11:25:00Z | Kalshi | December 1 | 99.5% | $169,301.53 |
| 2026-09-03T11:25:00Z | Kalshi | September 16 | 85% | $409,412.04 |
| 2026-09-03T11:25:00Z | Kalshi | October 1 | 88.5% | $409,412.04 |
| 2026-09-03T11:25:00Z | Kalshi | November 1 | 92.5% | $409,412.04 |
| 2026-09-03T11:25:00Z | Kalshi | January 1 | 93.5% | $409,412.04 |
ACCUMULATING HISTORY
Market movement, preserved
Each line is a market-implied price for one outcome. Two captures are enough to show the first move; a 7-day change stays blank until the dataset earns it.
Polymarket · OpenAI's Astra released by...?
History available| Outcome | Now | 24h | 7d |
|---|---|---|---|
| August 31 | 2% | — | — |
| September 15 | 97.35% | — | — |
| September 30 | 98.75% | — | — |
| October 31 | 98.85% | — | — |
| September 3 | 84.5% | — | — |
| September 4 | 91% | — | — |
| September 9 | 93.75% | — | — |
| September 11 | 96.15% | — | — |
| September 25 | 98.2% | — | — |
Kalshi · When will OpenAI release Astra?
First capture only| Outcome | Now | 24h | 7d |
|---|---|---|---|
| September 18 | 97.5% | — | — |
| October 1 | 98.5% | — | — |
| October 16 | 98.5% | — | — |
| November 1 | 99.5% | — | — |
| December 1 | 99.5% | — | — |
Kalshi · When will OpenAI release GPT-6?
History available| Outcome | Now | 24h | 7d |
|---|---|---|---|
| October 1 | 88.5% | — | — |
| November 1 | 92.5% | — | — |
| January 1 | 93.5% | — | — |
| September 16 | 85% | — | — |
Not official probability. Not trading advice. History is appended from public observations and API retrievals; missing liquidity is shown as missing.
RUMOR MILL, WITH RECEIPTS
Astra rumor tracker
Repetition is not corroboration. Each item below is traced to an upstream source and labeled for what the evidence can actually support.
Astra would launch the following week or on a specific August date.
The claim was traced to an anonymous X timing post. OpenAI's September 1 update says Astra will be available soon but still supplies no exact launch date, and no named reporter or independent confirmation supplies one either. The date is not safe to publish as a forecast.
Why it matters: It is a useful example of how a precise rumor can outrun the public record.
Upstream source: @synthwavedd on X ↗ · Fact-check review ↗
A “couple of weeks of releases” was treated as an Astra launch hint.
The post does not name Astra, specify a release date, or establish public access. It is an inference layered onto a vague release comment, not first-party confirmation.
Why it matters: Vague timing language becomes misleading when repeated as a model-specific announcement.
Upstream source: @thsottiaux on X ↗
Astra is definitely GPT-6.
The discussion combines the Astra name, GPT-6 market language, and assumptions about OpenAI naming. OpenAI has not confirmed that identity; reporting says the commercial label remained undecided.
Why it matters: A GPT-6 market and an Astra market have different resolution rules and cannot be merged.
Upstream source: Community timing discussion ↗
Astra itself breached Hugging Face during the earlier evaluation incident.
OpenAI names GPT-5.6 Sol and another pre-release model in the incident record and separately discusses Astra’s cyber-capability threshold. The public record does not identify Astra as the breaching model.
Why it matters: Safety reporting must not turn adjacent model incidents into an unsupported Astra claim.
Upstream source: OpenAI incident response ↗
OpenAI intends to make Astra broadly available but needs more time to do so safely.
The post is first-party context about intent and safety work, not a launch announcement. It gives no date, access path, API identifier, or price.
Why it matters: It is evidence about direction, not evidence that public availability is imminent.
Upstream source: Sam Altman on X ↗
Rumors checked September 5, 2026. This ledger records claims; it does not endorse them.
What to watch next
The next meaningful evidence would be an OpenAI product announcement, model documentation, public ChatGPT access, or an API model identifier. Until then, use the availability tracker for the access record and release-date timeline for the official chronology.
Questions people ask
What are the current Astra release odds?
At the latest capture (September 3, 2026 at 11:25 AM UTC), Polymarket showed 84.5% shown for release by September 3 and 91% shown by September 4. Kalshi showed 97.5% shown for its separate Astra contract before September 18. Market prices move and are not an OpenAI forecast.
Does a GPT-6 prediction market measure Astra?
No. Kalshi’s related series asks whether a public model whose name starts with GPT-6 will be released. It does not establish that GPT-6 is Astra, so its prices are a related naming signal rather than an Astra probability.
What would count as an official Astra launch?
For this publication, an official launch requires a current OpenAI announcement or product/API documentation showing public access. That threshold is now met by OpenAI’s September 3 safety overview, model documentation, and system card. Market movement remains a separate signal.
Is Astra launching next week?
The release date is confirmed as September 3, 2026; the remaining timing question is when access reaches each plan, account, and region.
ASTRA UPDATES
Keep the useful Astra updates coming
We’ll send occasional source-backed updates as access, pricing, and real-world tests change. This is an independent list, not an OpenAI waitlist.