Viete Studio / Models
Any Model. Your Choice. Your Control.
Model choice is configuration, not architecture. Choose among the models enabled in your deployment, match them to the task, and swap when the evidence supports it.
Browse the models enabled in your deployment. Each entry shows provider, endpoint, context capacity, and output limits, so you can match capability to the task.
Make selected models available to an agent and choose a default. Different agents can use different models, matched to what the work actually requires.
Agent-specific model choices
Assign one or more models to each agent. The agent uses the default you set, but can access alternatives when the task calls for different trade-offs.
Set limits per assignment
Configure context-window and output-token limits for each assignment. Keep resource use proportional to the task.
Display names and descriptions
Give each model assignment a name and description that makes sense to your team. Keep the interface clear even when multiple models are available.
Tasks differ in reasoning needs, response time, context size, and cost. Keep those trade-offs explicit rather than tying every agent to the same model.
Use recorded cost, token usage, and operational results to inform model choices. Evaluate in the context of your work, not against a generic benchmark.
Review cost and usage
Inspect run costs and token usage per model and agent. Identify where a different model could deliver the same result at lower cost.
Test in your own context
Run the same task with different models and compare results. Your requirements, not a leaderboard, determine which model fits.
Keep model choice a living decision
Model assignment is configuration, not architecture. Change it when the evidence supports a change, and verify the result before relying on it.