Choose a model
The model supplies the assistant's understanding, reasoning, and writing abilities. Start with the workspace default, then compare alternatives for your actual tasks.
On this page
When to switch
Straightforward extraction and formatting may work well with a faster, less expensive available model. Complex analysis or multi-step work may benefit from another model suited to that task.
For image tasks, choose a model that actually supports image input. A name alone does not establish its capabilities.
Change the session model
Use the model selector in the session. A change applies from the next model request; it does not rewrite previous replies or automatically rerun the whole task.
For a fair comparison, give two models the same materials and requirements. Compare accuracy, response time, output quality, and actual usage.
Missing or unavailable choices
Availability depends on platform supply, organization settings, user permissions, and current model status. Enterprise administrators can restrict model access.
Models and prices change. Treat the current console as the source for available choices and rates, rather than memorizing a fixed list from a guide.
Long conversations
Models have a limited capacity for content in one request. The platform may summarize older context in long sessions. A summary is not a guarantee that every detail remains available.
Save durable rules and conclusions in a cloud-drive document. Supply relevant materials instead of repeatedly including large unrelated files.
Custom models and Ollama
Technically supported teams can configure their own model services, including supported Ollama connections. This requires reachable services and correct capability and network settings.
Ordinary users do not need this setup. A custom model does not make all cloud tools, storage, and external services free; ask an administrator to investigate configuration errors.