OpenAI began making its newest large language model, GPT-6 Astra, available on September 3, but access isn’t as simple as clicking a button in ChatGPT. The model is currently accessible only to users on specific paid plans, including the $100 and $200 Pro tiers, as well as Business and Enterprise accounts. Free users and those on the Plus plan can access Astra through specialized tools like ChatGPT Work and Codex, but not through the standard ChatGPT interface. OpenAI is rolling out Astra gradually, so even eligible users may not get immediate access. For users on Pro, Business, or Enterprise accounts, Astra can be selected in the model picker within a regular ChatGPT conversation by choosing GPT-6 Pro from the Pro section. Enterprise users may need additional permissions set up in their workspace. Plus users can access Astra through the Work interface on web or mobile apps, but the desktop app requires switching from Chat to Work mode. Codex users must ensure their command-line interface is updated to version 0.153.0 or newer. If the desktop app hasn’t updated automatically, OpenAI recommends manually checking for updates via the app menu. Astra is being rolled out separately across different platforms, meaning it may become available on one platform before others. For users using Work with local files or desktop apps, it’s important to note that messages and task context might still be stored in the cloud, even when Work runs locally. This privacy consideration is something to keep in mind when using the model in these environments. Once available, users should be mindful of their usage limits, as Astra consumes tokens—measures of input and output—faster than the previous GPT-5.6 Sol model. Plus users can send about 5 to 45 local Astra messages every five hours, while the $100 Pro plan allows 25 to 225 messages and the $200 plan allows 100 to 900 messages. These limits reset periodically and can be tracked in the ChatGPT usage dashboard. The standard Chat interface has separate limits, with the $100 Pro plan allowing 50 GPT-6 Pro messages per week and the $200 plan allowing 200 messages per week. To avoid quickly exhausting their token allowance, OpenAI recommends reserving Astra for tasks that require its advanced capabilities. Users are advised to keep prompts concise, include only relevant files, define the desired output format and length, and specify any reasonable assumptions Astra can make to minimize unnecessary interactions. For developers, GPT-6 Astra is also available through the API, with a standard rate of $10 per million input tokens and $50 per million output tokens. Higher usage may trigger increased rates, depending on the number of tokens processed.