In short
- One model, every answer.A pinned model answers every question for that agent, so quality stays consistent.
- Chosen per agent.Keep support on Auto and give your sales agent a premium model.
- Cost before you commit.Each model shows its typical credits per answer in the picker.
- Backup built in.If the pinned model fails to start, another suitable model steps in.
Auto is the right choice for most agents. But some conversations deserve the strongest model you can get: a high-value sales question, a complex product, a brand that cares about every word. Pinning lets you choose exactly which model answers for a given agent.
How to pin a model
- Open the agent and go to its Model tab.
- Compare models by tier, credits per answer and what each is best for.
- Click Use on the model you want.
- The agent answers with that model from the next question.
You can go back to Auto at any time with one click. Every agent has its own setting, so you can mix models across agents in the same workspace.
Which plans include which models
Pinning a model is part of every Premium size and Enterprise, and covers every model, including the Ultra models, the largest Claude and GPT models. Free cannot pin a model and uses the fast and standard models.
Premium with your own AI key is coming soon. You will connect a key from Anthropic, OpenAI, Google Gemini, OpenRouter or Moonshot, and every answer will count as one answer, whatever the model.
What a premium answer costs
Stronger models cost more to run, so they use more credits per answer. The figure in the picker is a typical estimate. Each answer is billed by the tokens it actually used, and your analytics show credits by agent and by model, so you always see what a pinned model is using.
Safe when things change
If a pinned model does not start answering in time, OneType moves on to another suitable model so the visitor still gets an answer. If your plan changes and no longer includes the pinned model, the agent switches to Auto on its own instead of going silent.
When a pinned model makes sense
- High-value sales conversations where tone and detail matter.
- Complex or technical products with long, detailed content.
- Brands that want the same model behind every answer.
- Agents that handle questions with many parts.
Whatever model you pin, the rules stay the same. The agent answers from your content, cites its sources and says it does not know when your content has no answer. A stronger model writes better answers; it does not change what your agent is allowed to say.
Try before you commit
Because each agent has its own setting, you can test a stronger model on a copy of your agent first. Duplicate the agent, pin the model on the copy, and compare answers to the same questions. When you are happy, pin the model on the live agent, or keep Auto if the difference does not justify the extra credits.