Skip to main content
Dume doesn’t lock you into a single model. Switch models per message from the composer, or set a default in your profile. The model you pick changes two things: how the response feels, and how many credits it costs.

​
Available models

The picker groups models into two lists:
GPT 5, GPT 5.1, GPT 5.4, GPT 5 Mini, GPT 5 Nano, GPT 5.3 Codex, GPT 5.6 Luna, GPT 5.6 Terra, GPT 5.6 Sol
Ternary Bonsai 27B is always free, on every plan, no credit cost per message. It’s the default model for free-plan accounts. Paid accounts default to Kimi K2.6. Every other model is priced individually and costs credits per message based on actual usage (see below). Which of them you can select depends on your plan: anything outside your plan shows a lock icon in the picker.
Free plans have access to a smaller set of models. Anything locked unlocks by upgrading; see Billing & plans.

​
How model choice affects credits

Dume doesn’t charge a flat fee per message. Every model has its own credit cost, and you’re billed for what a message actually uses: a short question to a lighter model costs a fraction of a credit, while a long, complex request to a premium model costs more. Ternary Bonsai 27B costs nothing at all, on any plan. Model names like “Nano,” “Mini,” and “Codex” are a rough guide to what each is tuned for; if you’re trying to conserve credits, default to a smaller model for routine questions and reach for a premium one when the task actually needs it. See the Credit Cost Guide for approximate costs on common tasks.

​
Setting a default

Go to Settings → Profile → System Preferences and set your Default AI Model. New sessions start with this model unless you change it in the composer.