This is the place to search out GPT-6 Astra and what to know earlier than its utilization limits catch you off guard.
OpenAI started rolling out GPT-6 Astra on September 3, however entry is just not as simple as opening ChatGPT and trying to find a button that claims “Astra.” In common ChatGPT, the mannequin is named GPT-6 Professional and is rolling out for the $100 and $200 Professional plans, Enterprise, and Enterprise. For now, Free and Go customers are excluded, with Plus subscribers getting Astra by way of ChatGPT Work and Codex as an alternative (not common Chat) because the rollout reaches their accounts.
Availability might fluctuate between Chat, Work and Codex. OpenAI says the rollout might be gradual, so being on an eligible plan does not imply Astra might be ready for you in the present day. OpenAI hasn’t specified when eligible accounts will obtain Astra.
Tips on how to use GPT-6 Astra in ChatGPT
If in case you have an eligible Professional, Enterprise or Enterprise account, open the mannequin picker in a daily ChatGPT dialog and look beneath Professional for GPT-6 Professional. That is the one powered by GPT-6 Astra, so do not get too caught up within the menu in search of its flashier identify. Enterprise customers have one other potential impediment: entry additionally is dependent upon workspace model-access permissions.
When you’re on Plus, open Work on the net or cell app and choose Astra from the accessible fashions as soon as it rolls out to your account. Within the desktop app, choose ChatGPT, change from Chat to Work and select Astra there. Codex customers may select Astra, however OpenAI says the Codex CLI requires model 0.153.0 or newer.
Nonetheless nothing? Even when the desktop app has just lately up to date itself, OpenAI suggests manually checking for updates by going to Menu > Test for Updates. Work is rolling out individually, so Astra is perhaps accessible on one platform earlier than one other.
In case you are utilizing Work with native recordsdata or desktop apps, there’s additionally a privateness side. Even when Work runs domestically, messages and job context should be saved within the cloud.
Watch how rapidly Astra makes use of your allowance
When Astra does arrive, there’s one other issue to regulate: your utilization allowance.
In keeping with OpenAI, Astra can dissipate your allowance extra rapidly than GPT-5.6 Sol, relying on the complexity of the duty, the quantity of context, the reasoning required and different utilization settings. So a giant coding task that requires extra reasoning can use extra of your allowance than a fast query.
OpenAI estimates that Plus customers can ship roughly 5 to 45 native Astra messages per five-hour interval. The $100 Professional plan is estimated at 25 to 225 messages, whereas the $200 tier will get 100 to 900. You may test your present limits and reset instances in ChatGPT’s utilization dashboard.
Chat has separate limits. On the $100 Professional plan, GPT-6 Professional shares a 50-message weekly allowance with GPT-5.6 Sol Professional, whereas the $200 plan permits 200 GPT-6 Professional messages per week.
In case your allowance goes down faster than you need, reserve Astra for the duties that require it. OpenAI advises maintaining prompts centered, together with solely related recordsdata and defining the output format and size. Telling Astra what cheap assumptions it may make might also cut back pointless back-and-forth.
GPT-6 Astra is obtainable for builders by way of the API. The usual value is $10 per million enter tokens and $50 per million output tokens. Requests that rack up greater than 272,000 enter tokens in complete are charged at twice the enter and cache charges, and output charges at 1.5x.