Update: GPT-6 Astra Pro is callable per-token, not just in ChatGPT
OpenRouter now serves Astra's 'pro' reasoning tier at $10/$50, 1.05M context — same weights, dialed up. Who gets it in ChatGPT, and what OpenAI won't say.
Copy markdown
The 'pro' tier is now a per-token API model
On OpenRouter, `openai/gpt-6-astra-pro` went live September 4 at $10 in / $50 out per million tokens — identical to standard Astra, with the same 1.05M-token context and 128K max output. It's the same weights served with `reasoning.mode: pro`, so you get Astra's highest-quality answers on hard tasks by switching one model ID, no ChatGPT subscription needed.
In ChatGPT, Pro is gated to the paid plans
Inside the app, Astra Pro is rolling out to ChatGPT Pro, Business, and Enterprise first; Plus stays on standard Astra and there's no free-tier date. Reports describe the in-app Pro tier as subscription-only and "not sold per token" — which sits awkwardly beside the live per-token OpenRouter listing.
Same sticker price — cached input is the real lever
Astra and Astra Pro both bill $10 input / $50 output per million, but cached input is just $1 per million. For agent loops that resend a big system prompt every turn, that 10x cache discount moves your bill far more than the reasoning tier does — structure for cache hits before you reach for Pro.
The catch: OpenAI won't say what 'pro' improves
OpenAI has published no separate benchmarks, context limits, or capability specs for the Pro configuration — so there's no way to quantify the gain before you pay for it. Treat it as a quality dial to A/B on your own evals, not a documented step up.