Choosing a model

The model picker sits below the chat box. Most people never need to touch it.

The three models

  • GPT-6.1 Sol (recommended). The default for everything: prospecting, drafting, research, browsing funder portals. In our testing it was the most accurate writer.
  • GPT-6 Astra (for hard tasks). The strongest and slowest. Use it for real judgement calls, like a tricky eligibility question or a confusing funder requirement. It uses about 3x the usage of Sol, so a task that takes 100 on Sol takes about 300 here.
  • Claude Opus 5.5 (an alternative). Some people prefer its writing style. It's more willing to fill a gap with a plausible fact or number, so check its figures. It uses about 2x the usage of Sol in a chat, and about 4x on a scheduled run.

Claude does everything Sol does in GrantLoop. Big research jobs and column fills still run on GPT. If Opus is busy, the step goes to Claude Sonnet 5.5, then to Sol. When you pick Claude, your conversation and the files it reads go to Anthropic, the same way they go to OpenAI for GPT. Neither provider trains on it.

Switching

  • In chat: click the model name below the message box and pick one.
  • Each chat remembers its model, even on another computer. Other chats are unaffected.
  • A new chat opens on the model you picked last, or Sol if you haven't picked one. If you switched to a pricier model for one job, switch back when you're done.
  • GPT and Claude chats stay separate. You can move between Sol and Astra inside a chat. Picking Claude in a GPT chat, or GPT in a Claude chat, opens a new chat (the picker marks these "new chat").
  • Switch at the start of a task if you can. The first message after switching between Sol and Astra re-reads the whole chat, so in a long chat it can cost 20 to 50 credits. After that it's back to normal.
  • For a scheduled run: choose a model in its settings. The change applies from its next run. Runs on a pricier model show a small tag in your list.
  • Or just ask: "Set my Monday grant sweep to use Claude Opus."

Coming from an older model

Chats and scheduled runs on GPT-6 Sol, GPT-5.6 Terra or GPT-5.6 Sol now run on GPT-6.1 Sol. You don't need to change anything. If GPT-6.1 Sol is busy, the step goes to GPT-6 Sol.

Which should I use?

Start on Sol and stay there. Switch to Astra for something genuinely hard, then switch back. Leaving a busy scheduled run on a pricier model is the fastest way to use up a month's allowance.

If you regularly run out, a bigger plan usually beats a smaller model. Get in touch and we'll help you size it.