OpenAI began rolling out GPT-6 Astra on September 3 — but what you actually get depends heavily on which ChatGPT plan you're paying for. Plus subscribers see a model explicitly named GPT-6 Astra, available through ChatGPT Work and Codex. Pro, Business, and Enterprise customers instead get something called GPT-6 Pro, which OpenAI says is "powered by" Astra — not the same named product.
That's a meaningfully different rollout than the initial announcement implies. This isn't one uniform model upgrade landing everywhere at once — it's two differently branded offerings tied to the same underlying system, and OpenAI's documentation doesn't establish whether Astra and GPT-6 Pro actually expose the same features, limits, or behavior in practice.
For Plus users specifically, Astra is showing up in two named surfaces: ChatGPT Work and Codex. Access to Work itself is still expanding to eligible accounts, meaning a valid paid subscription doesn't guarantee you can use it right now — access depends on your plan, the specific product surface, and your account's rollout status.
The bigger gap is on performance, not access. OpenAI hasn't published benchmark methodology, reproducible evaluations, architecture details, or task-level comparisons against earlier models. So while the rollout itself — who gets what, and where — is well documented, there's currently no public evidence for improved coding reliability, tool use, latency, pricing, or safety behavior.
Bottom line: GPT-6 Astra's rollout is real and specific: Plus gets "Astra" in Work and Codex, higher tiers get a related "GPT-6 Pro." What isn't established yet is whether either version is actually better in practice — that evaluation hasn't been published.
