OpenAI initiated the deployment of its latest language model, GPT-6 Astra, on September 3. While the technology represents a significant leap in intelligence and alignment, accessing the model is not as straightforward as searching for an “Astra” button within the standard ChatGPT interface. Under the hood, the model appears in regular ChatGPT conversations labeled as GPT-6 Pro, specifically targeting users on the $100 and $200 monthly Pro subscriptions, as well as Business and Enterprise accounts.
Currently, free users and those on Go plans are excluded from access. Plus subscribers may gain entry through ChatGPT Work or Codex platforms rather than the standard chat interface, depending on how the rollout reaches their specific accounts. OpenAI has characterized the distribution as gradual, noting that subscription eligibility alone does not guarantee immediate availability. The company has not released a specific timeline for when eligible accounts will receive the update.
For users with eligible Pro, Business, or Enterprise credentials, the model can be selected via the model picker in a standard ChatGPT conversation under the “Pro” category. Enterprise users must also ensure their workspace administrator has granted the necessary model-access permissions. Meanwhile, Plus subscribers should navigate to the Work interface on web or mobile, where Astra will appear in the model selection once the account is eligible. In the desktop application, users must switch from the standard Chat view to Work to select the model. Codex CLI users require version 0.153.0 or newer to utilize Astra.
Should the model remain inaccessible despite recent app updates, OpenAI recommends manually verifying for updates through the Menu and selecting “Check for Updates.” Because the Work feature is rolling out independently, availability may vary across platforms. Additionally, users managing local files or desktop apps within Work should be aware that while processing occurs locally, message history and task context may still be stored on OpenAI’s cloud servers.
A critical factor for users is the consumption of their usage allowance, which depletes significantly faster with Astra than with the previous GPT-5.6 Sol model. Consumption rates depend on task complexity, context length, and reasoning requirements. OpenAI estimates that Plus users can send between five and 45 Astra messages within a five-hour window. The $100 Pro plan allows for an estimated 25 to 225 messages, while the $200 tier permits 100 to 900 messages in the same period.
Message limits are tracked separately for Chat. On the $100 Pro plan, GPT-6 Pro shares a weekly cap of 50 messages with GPT-5.6 Sol Pro. The $200 plan provides a distinct allocation of 200 GPT-6 Pro messages per week. To conserve allowance, OpenAI advises users to reserve Astra for complex tasks, keep prompts concise, attach only necessary files, and clearly define output formats and reasonable assumptions to minimize iterative back-and-forth.
Developers can access GPT-6 Astra via the API, priced at $10 per million input tokens and $50 per million output tokens. Large requests exceeding 272,000 input tokens incur higher rates, with input and cache charges doubling and output charges increasing by 50%.
Finally, some hope for free users! But honestly, the pricing tiers look like a classic upsell trap for the Pro plan.
The usage caps on the $100 Pro tier seem brutal. I wonder if they’ll adjust limits as users get used to the consumption rate.