Explore different minds for different kinds of work.
A service you want to launch needs research, an offer people can understand, and a first conversation with customers. Those jobs ask different things of an autonomous workforce. Supanova lets you choose from supported models and providers, then make those choices available to the specialists carrying your work.
Multi-model is the starting point for shaping that intelligence. You decide which models your workspace may use. From there, you can assign models to parts of the workforce and, for supported work, choose whether model selection should learn from completed results. The ambition stays yours; the technology behind each responsibility becomes a choice you can examine.
Imagine you want to help independent shops bring their shelves online. Before you write an announcement, you need to understand what shop owners struggle with, decide which part of the job your service will handle, and describe the offer in words they would trust. Interview notes need sorting; conflicting answers need a careful reading; the offer needs a clear argument.
An AI specialist can take on each part as a defined task or as connected work within a project. The right model for extracting useful facts from source material may differ from the one you choose for difficult synthesis or writing. Supported models bring different capabilities, context capacity, tool support, speeds, and costs to those jobs.
In this illustrative project, a research specialist organizes shop-owner interviews and flags a question the notes have not settled: are product photos or inventory entry the bigger obstacle? Another specialist prepares an offer outline from the approved findings. A writer drafts an introduction. You review the evidence and decide which part of the service you can promise. The model choices support a connected result you can judge.
Offer under review: Begin with a small online catalog of the items customers ask for most. Open decision: Are product photos included in the service or supplied by the shop?
In your workspace's Multi-Model settings, you can enable supported model providers and individual models. That creates the palette your workforce is allowed to draw from. The catalog shows available capabilities and cost indicators, so you can choose with the work in mind.
The palette makes models available for configuration. If you enable a second model for more demanding research, your existing workforce assignments stay in place until you change them. You can start with a focused set that fits one workflow and widen it when another model has a clear job to do.
Availability depends on the models and providers Supanova supports and on your workspace's configuration. The current workspace catalog is the place to check what you can actually use. A model that looks interesting elsewhere may be unavailable or unsuited to a task's required capabilities.
Once the palette is set, Model controls lets you choose model assignments for workforce bands and, where appropriate, finer tier settings. That is how the launch project can give routine preparation and more complex coordination different configured defaults. The brief, sources, and approval standard still tell each specialist what good work looks like.
Judge a model choice by what comes back. Did the research separate evidence from guesswork? Did the offer need several revisions before you could use it? Was the customer introduction clear without making a promise you had not approved? Compare the cost of reaching an accepted result along with the quality of the first attempt. A low-cost first pass can be expensive if you keep repairing it; a capable model is only valuable when its work helps the project reach a useful finish.
You can make a deliberate change, inspect comparable work, and keep the assignment that serves your standard. Model choice becomes part of how you direct the workforce.
Dynamic Model Selection is a separate, paid workspace add-on for supported tasks. If enabled, it can choose among models your workspace permits, within the task's capability needs and your quality preference. Your configured model remains the baseline while the selector gathers enough evidence from completed work; it also remains a fallback when evidence or selection is insufficient.
The palette defines what is available; Dynamic Model Selection decides among eligible options for supported tasks. Adaptive Compute adjusts how much thinking headroom a selected model receives. Each has its own setting, and your compute and spend limits remain in force.
Can one project use more than one model?
Different specialists can contribute to a connected project under the model assignments configured for their workforce tiers. Start with the approved palette, then choose the assignments that fit the work. Review dependent assignments before removing a model from the palette.
Do I need to know every model before I begin?
No. Begin with a responsibility you can judge and the supported models available in your workspace. Compare capabilities, revision effort, and cost against that work. Expand your palette when a new model has a specific purpose.
How will I know whether another model helps?
Give comparable work the same brief and acceptance standard. Look at the first result, the revisions needed, and the cost of work you can use. A model has earned a place in your palette when it helps a responsibility meet your standard.
Choose a project worth moving and the evidence you will use to judge it. Supanova can put a workforce behind the steps; you can decide which supported models belong in that workforce as you learn from the results.
Launch a new workspace
Previous: Agent progression · Next: Multiplayer · Back to the Feature Catalog
Launch a new workspace