PromptQL works with a range of leading language models, so you can pick the right one for each task — and switch whenever you like.

• Proprietary models from providers like Anthropic (Claude), OpenAI (GPT), and Google (Gemini).
• Open-weight models (for example, served via Fireworks) when you prefer open models or want to optimize for cost.

You choose the model per thread with the model picker. You can also set a default for the project or for a room.

Most models are available out of the box — just pick one in the model picker.

If you have an OpenAI Codex plan (a ChatGPT subscription), you can connect it and use GPT models through your own plan instead of paying for tokens with PromptQL credits.

To connect it:

1. Click on the user card at the bottom of the sidebar.
2. From the menu, choose Profile.
3. Next to Use your Codex plan, click Add ChatGPT token.
4. Paste your ChatGPT access token, obtained from chatgpt.com/admin/access-tokens.

Note: An admin needs to enable creating access tokens on your ChatGPT Workspace.

Once connected, GPT models run via your own Codex plan, so you're not spending PromptQL credits on tokens for that model.