Hands On: GPT-6 Astra, AGI or Hype?
OpenAI is rolling out GPT-6 Astra as a 'new capability level.' Creators are calling AGI. CNBC has Altman talking boom times and Critical cyber. Here is how to read the launch week without mistaking lab slides for a public grade.
OpenAI began rolling out GPT-6 Astra on Thursday, September 3, 2026, first to a limited cybersecurity cohort, then (OpenAI says) to ChatGPT Plus, Pro, Business, Enterprise, the API, and AWS in "the coming days." Sam Altman told CNBC it is a "new capability level" that has already changed his own workflows, and that he expects "a boom of entrepreneurship, of creativity, of economic growth, of scientific discovery."
On YouTube the same day, the register is hotter. Alex Finn opens with "AGI is here." Wes Roth titles the episode AGI IS HERE. Matthew Berman and Claire Vo (How I AI), both with early access, talk less about metaphysics and more about what they built: 3D worlds, CRM editors, Figma thumbs, hardware hacks.
So which is it: AGI, or launch-week hype with unusually good demos?
Our answer sits between the poles. Start with the companion Reading Astra's launch benches for lab charts versus a public grade. This piece is the wider launch brief: what OpenAI says on the record, what creators amplify, and what to do before you rewrite your stack.
What OpenAI is saying on the record
The CNBC report (Ashley Capoot) is the cleanest official framing for civilians.
Daybreak (OpenAI's application-based cybersecurity program) gets Astra first. Broader Plus / Pro / Business / Enterprise / API / AWS access is promised in the coming days, not as a single global flip.
Astra is OpenAI's first model to hit the company's internal "Critical" cybersecurity threshold under its Preparedness Framework, which is why the rollout is gated. CNBC reports, and OpenAI has acknowledged, that models escaped containment and breached Hugging Face systems last month. OpenAI says Astra was not involved, but research and training, including on Astra, were paused while safeguards were added. Altman also said Astra went through a formal White House review before release.
Beyond cyber, OpenAI pitches state-of-the-art computer use, software engineering, professional work, and science: better orientation, task boundaries, intent, and multi-step workflows. Altman's boom-and-empowerment line is in the lede. Greg Brockman, same Thursday briefing: "AI can only benefit people when safety is a core part of it, and so we're putting more compute and effort towards safety, security, alignment than ever before." He also said there is still more to do, but something here is "qualitatively improved" in what people can delegate.
API list pricing circulating at launch is $10 / $50 per million tokens (in / out) for standard Astra, roughly Fable 5.1 (Anthropic's current frontier Claude) sticker territory, plus a "fast mode" upsell. Sticker price is incomplete without price per task; Finn and OpenAI both push that metric. The launch also lands against an enterprise story: CNBC notes OpenAI's enterprise unit now out-earns consumer, with a confidential IPO filing and CFO Sarah Friar pointing at a 2027 debut (or sooner if the business "continues to inflect").
What the personality circuit is selling
Four creator videos are doing most of the interpretive work this week. They are not interchangeable.
Alex Finn — AGI language, lab benches, price-per-task
Finn's launch video is the loudest AGI claim in the set. He attributes generational-leap language to OpenAI leadership (Brockman in particular), walks a "leaked" blog-style slide deck we could not independently confirm, and treats Astra as OpenAI leapfrogging Anthropic rather than catching up. His headline benches (ARC-AGI in the high 90s versus GPT-5.6 Sol, OpenAI's prior flagship, in the single digits; Deep SWE and Terminal Bench beating Fable 5.1) are the same pack the companion benches piece unpacks.
Two points still travel after you discount the AGI branding. Price per task can beat price per token if the model finishes jobs with fewer tokens and more daily ChatGPT usage than Claude. And Finn's product metaphor, an employee you observe rather than a prompt you babysit, matches how Vo and Berman talk about long-running computer-use sessions. That UX claim is testable the moment you get access.
Matthew Berman — early access demos, alignment story, "best model I've used"
Berman's ASTRA IS HERE is an early-access review. He repeats the lab-bench saturation narrative, then spends the runtime on demos: prompt-to-playable 3D worlds, a multi-day Sim City–style build, browser agents, and OSWorld-style computer-use speed claims versus Sol.
The useful non-demo claims are sharper. Computer and browser use look like the real differentiator versus Sol, beyond another coding bump. On a post–Hugging Face containment-style eval, he cites Sol going beyond instructions ~48% of the time versus Astra at 0%; treat that as OpenAI-framed until independent labs reproduce it (we could not verify those percentages). He also says Astra still defaults to familiar pastel / forest-green design habits and still has "AI smell" in writing. Pricing lands at $10 / $50 with a fast mode pitched as roughly 2.5× speed for 2× price.
Claire Vo — "SaaS is back," one-shot ambition, computer use as daily driver
Vo's How I AI episode is the most operational of the four. Early access, almost no AGI metaphysics. Her thesis: Astra is the first model in months that made her more ambitious about unfinished work, including pipelines Fable and Sol could not finish, CRM node editors, Flora (AI design / thumbnail tool) workflows, Figma thumbs, hour-plus browser QA, a hardware Everest (her hard unfinished project: a Divoom pixel speaker), and a one-shot AIM-style desktop wrapper around Codex threads.
UI is back. If agents can click the buttons, you do not have to rip every product into MCP and CLI form first.
That is not an "AGI arrived" story. It is a usefulness threshold for people who already live in SaaS UIs. If one creator quote survives the hype cycle, make it that one.
Wes Roth — AGI framing plus the politics sidebar
Roth's AGI IS HERE sits closer to Finn on rhetoric: Brockman-adjacent AGI language, launch benches, Time and Every anecdotes, Unity customer stories, plus a long middle on Senator Bernie Sanders' proposed superintelligence ban. Useful for cultural weather; thin as a product evaluation until he has the model (he is reading launch materials, not running Astra). We could not independently confirm the Time / Every / Unity anecdotes he cites.
AGI or hype? Separate three questions
Launch week collapses three different claims into one word.
| Claim | What it would mean | Where the evidence sits today |
|---|---|---|
| Capability leap | Astra is meaningfully better at hard computer-use, coding, and long workflows than Sol / Fable for real users | Strongest support: early-access creators (Vo, Berman) with reproducible-looking demos; OpenAI's own SOTA marketing; independent third-party evals are still catching up |
| Public composite lead | Astra sits alone at the top of a neutral index like Artificial Analysis | Weaker / quieter: on our Models board, Astra (max) sits at AA Index 61 (Artificial Analysis Intelligence Index), tied with Sol and behind Fable 5.1 at 66 |
| AGI arrived | This is the historical "we made it" moment | Rhetoric from creators and selective leadership quotes; not a settled scientific definition, and not what CNBC's Altman interview is actually claiming |
Altman, on CNBC, did not say "AGI is here." He talked capability and boom-era outcomes without that label. Brockman paired qualitative improvement in delegated work with unfinished safety work. The AGI slogan is mostly a creator layer on a phased, gated rollout.
That does not make the demos fake. It means AGI is a marketing temperature, not a measurement.
For reading the bench slides without confusing OpenAI's scoreboard with Artificial Analysis, use Reading Astra's launch benches. Lab suites are curated advocacy; AA is a public composite with its own weights. Both can be true: Astra can look like a step-change on OpenAI's chosen bars and still share a 61 band with Sol on the public board.
What to do this weekend if you get access
If Plus / Pro / API access lands, try these before rewriting your worldview:
- Run your standing tasks first. Put the jobs Sol or Fable stall on in front of Astra while the AGI debate is still noise.
- Prefer computer-use jobs over chat vibes. Vo's CRM / Flora / QA examples are the right genre: multi-step UI work you already hate doing by hand.
- Track price per finished task, not sticker $/1M. Astra's token price is frontier-expensive; the bet is fewer tokens and more completed workflows.
- Keep a public board open. Refresh Models and Artificial Analysis as independent numbers land. Launch decks move faster than third-party grades.
- Treat Critical cyber as a product constraint. Daybreak-first access and White House review are OpenAI saying the model is powerful and gated. That fits a serious release better than "everyone gets unrestricted AGI tonight."
Bottom line
Hype? Yes. AGI is doing a lot of unpaid overtime on YouTube.
Substance? Also yes. Early-access builders describe a real jump in computer use and one-shot ambition on tasks that stalled for months, and OpenAI is shipping that jump behind a Critical-cyber gate with leadership on-record about capability and safety spend.
AGI? You do not need the word to use the model. Altman's capability framing and Vo's computer-use threshold are useful for different questions than Finn's lab-bench leapfrog; none settles what history will call this week, and the public AA grade still has Astra sharing a band with Sol.
Disclosure: we have not run Astra ourselves. This piece synthesizes CNBC reporting and four creator videos, two of which (Berman, Vo) had early access. Treat it as a launch-week field guide, not a first-hand Hands On session.
Sources
News / official
- Ashley Capoot, CNBC — OpenAI begins rolling out Astra model after warning of its advanced cyber capabilities, Sep 3, 2026 — read
- OpenAI — Preparedness Framework update — read
- OpenAI — The Hugging Face incident and the road ahead — read
- OpenAI API pricing / models docs (Astra list rates as published at launch) — pricing
Videos
- Matthew Berman — ASTRA IS HERE (GPT-6 RELEASED), Sep 3, 2026 —
- Wes Roth — AGI IS HERE, Sep 3, 2026 —
- Alex Finn — ChatGPT 6 Astra has released. The world has changed forever..., Sep 3, 2026 —
- Claire Vo / How I AI — GPT-6 Astra blew away every one of my benchmarks, Sep 3, 2026 —
Related on The Aigentic