Choosing a Model

GrantLoop runs on OpenAI's GPT-6, and gives you two versions of it to work with. You pick one from the model name below the chat box: click it and choose.

GPT-6 Sol: the default

For most tasks. Sol runs everything unless you say otherwise: prospecting, drafting, research, browsing funder portals, answering questions about your workspace.

If you never touch the model picker, you are on Sol, and that is the right call for almost everything.

GPT-6 Astra: for the hardest tasks

Astra is the stronger of the two. Reach for it when a task involves genuine judgement rather than volume: a tricky eligibility call, a nuanced piece of writing, untangling a confusing funder requirement, or reviewing work you already have.

Astra uses about 3× the credits Sol does for the same work, which is what the picker shows next to it.

That is the whole trade. A task that takes 100 credits on Sol takes about 300 on Astra.

There is no penalty in switching. You are simply spending your allowance faster in exchange for a stronger model. Use it where the extra judgement earns its keep, and leave it off for routine work.

Switching

In chat: click the model name below the message box and pick.

Each conversation remembers its own model. If you switch a chat to Astra, that chat stays on Astra every time you come back to it, including from a different computer. Other chats are unaffected, and a chat you never switched stays on Sol.

A new chat opens on whichever model you picked last, or Sol until you have picked anything. Existing chats never move: picking Astra for one hard question does not reach back into conversations you already had.

You can still switch mid-conversation whenever you want. It simply moves that conversation onto the new model from that point on.

Switch at the start of a task where you can. The first message you send after changing models has to re-read the whole conversation from scratch, so it costs noticeably more than the messages around it. A long chat can spend 20 to 50 credits on that one message. It settles back down immediately afterwards, and switching back later is cheap. Changing models in a fresh or short chat costs almost nothing.

For a recurring task: open the task and choose a model in its settings. A change applies from the task's next run; a run already in progress keeps the model it started with. A task set to Astra shows a small GPT-6 Astra tag in your task list, so you can see at a glance which ones are on the pricier model.

Just ask: you can also tell the agent directly. "Set my Monday grant sweep to use Astra" or "put that task back on the default model" both work.

Coming from GPT-5.6

Chats and recurring tasks that were on GPT-5.6 Terra or GPT-5.6 Sol now run on GPT-6 Sol automatically. You don't need to change anything, and nothing moved onto the pricier Astra without you choosing it.

Which should I use?

Start on Sol and stay there. It handles the large majority of grant work well, and it is what every new chat and every recurring task uses out of the box.

Switch to Astra when you have hit something genuinely hard and want a second, stronger opinion, then switch back. Because Astra spends credits about 3× faster, leaving a busy recurring task on Astra is the quickest way to burn through a month's allowance without meaning to.

If you are regularly running out of credits, the fix is usually a bigger plan rather than a smaller model. Get in touch and we will help you size it.