Regular Chat's model menu won't gain GPT-6.2 Astra or Sol in this rollout. Switch to Work in ChatGPT or open the model picker in Codex. If GPT-6.2 is missing there too, check your client, sign-in method, and workspace. OpenAI announced both models on November 17, 2026. The rollout and access rules below were checked on November 18, 2026. GPT-6.2 announcement · ChatGPT Work and Codex help
Open the model picker in a new task
| Where you are | What to check |
|---|---|
| Regular Chat | The GPT-6 and Pro options belong to regular Chat. Switch to Work to look for the two new models. |
| ChatGPT Work | Start a new task with the same account and workspace. Open the model picker. If you see the Power slider, expand Advanced to check the available models. |
| Codex in the ChatGPT desktop app | Switch to Codex, start a new thread, and open the model picker below the composer. |
| Codex CLI | Run codex --version to see which version your terminal is running. Once the CLI is up to date, start a fresh interactive session and enter /model to see the available choices. |
If you can select 6.2 Astra or 6.2 Sol by name in Work or Codex, that model is available in the picker you checked. Existing Work chats may keep their original model, and older Codex threads may retain a manual selection. If you only see Default or an old thread has no upgrade prompt, check the picker in a new task. Model selection · Model updates in Work and Codex
Update the client and compare the same workspace
In the ChatGPT desktop app, choose Check for Updates from the app menu, then restart after updating. If you use Codex CLI, update the copy your terminal actually runs and check codex --version again. If GPT-6.2 is missing in desktop Work, start a new task in Work on the web with the same account and workspace. If you can select 6.2 on the web, that confirms access in web Work. You can continue there while you investigate the desktop difference. If both menus are missing the model, check how you're signed in and what your workspace allows. OpenAI rollout guidance
Confirm which account and billing Codex uses
Open the desktop profile menu to check the active account and workspace or API key status. In the CLI, run codex login status from the terminal you use for Codex to check the active sign-in method.
- With ChatGPT sign-in, model access follows your ChatGPT plan and workspace permissions. Confirm that the active workspace is the one associated with your personal subscription or organizational seat.
- With API key sign-in, usage is billed through OpenAI Platform. Check the API project associated with the key and its model permissions. The API model IDs are
gpt-6.2-astraandgpt-6.2-sol. Use ChatGPT sign-in if you intended to use your subscription allowance.
In the CLI, codex --model gpt-6.2-sol selects a model for a new session. You can also check whether an existing config.toml pins an older model. Settings and flags specify the model you want to use. Successful startup or a model name in the config alone does not establish that a request will succeed. Authentication and billing · Codex model configuration · API model documentation
Check your plan, workspace permissions, and rollout status
- Personal Plus or Pro. Both 6.2 models are in the rollout for Work and Codex. Plus users can select the models when available on their account without upgrading to Pro. If the models are missing after an update, confirm which account and workspace hold the subscription. OpenAI release announcement · Plus vs. Pro guide
- Personal Free or Go. The launch announcement lists Plus and Pro as the eligible personal plans. Access to regular Chat or Codex alone does not establish that your plan can select GPT-6.2. OpenAI release announcement
- Business. OpenAI's Help Center includes Business in the phased rollout. Check the picker in your Business workspace and ask your admin to confirm your seat and workspace settings. A Premium seat alone does not tell you whether the models have reached your account. Business models and limits
- Enterprise or Edu. After the preview reaches a workspace, both models are off by default in Work and desktop Codex. An admin can search for a member in Admin console → Models → Test and inspect their effective model permissions and roles. If the models have reached the workspace and access is disabled, the workspace owner can enable access for the relevant members through workspace or role settings. OpenAI also states that workspaces requiring UAE inference cannot use these models in those experiences. Enterprise and Edu models and permissions
If you've confirmed the experience, version, sign-in method, and permissions, the rollout may still be pending for your account. OpenAI's help page gives no per-account schedule. Use an available model and check a new task later. If you contact an admin or support, include the experience, version, plan or seat, sign-in method, and model names shown in the picker. OpenAI rollout guidance
GPT-6.2 is selectable, but Ultrafast is missing
Ultrafast is a speed setting for an eligible model. If 6.2 is available but you can't select Ultrafast, check the model, where you're using it, and your plan.
| Plan | GPT-6.2 Ultrafast eligibility |
|---|---|
| Pro 200 (US$200/month) | GPT-6.2 Sol in Codex only. |
| Pro 500 (US$500/month) | Both 6.2 Astra and 6.2 Sol in Work and Codex. |
| Enterprise or Edu | Available for eligible workspaces with the required permissions. |
| Plus, Pro 100, Business | Unavailable, including with purchased credits. |
These rules apply when signed in to a ChatGPT workspace. For API usage, check the separate Ultrafast eligibility and rates. Work and Codex Ultrafast rules · API Ultrafast documentation