Meet your workspace.

VARIANT-1 is a Windows AI workspace for work that continues beyond a single answer. Your model can run Python, work with files, browse, and use supported desktop tools from the same chat.

Each chat has a live Python environment. Data and helper functions can stay available between calls, so a follow-up can build on the work already done.

Choose a starting point

From a request to a result

Describe what you want and which materials the model should use. It can combine code and tools, show its results, and respond to corrections. Files, browser views, a terminal, and review panes sit alongside the conversation.

You don’t need to write Python to give a task. The model writes and runs the code; its ability to complete the work depends on the model, tools, and inputs.

About the beta

The application source is available as an early preview under the MIT License. A free Windows installer is still in preparation. You supply inference through a supported local model or your own provider connection. No VARIANT account, subscription, or managed inference balance is required for this beta.

The public installer is in preparation. These guides describe the current development version; final setup steps, tested connections, and known limitations will accompany the release.

Keep the work under your control

Begin with clear scope and sample materials. Inspect important outputs and use steering or stopping when the direction needs to change. VARIANT-1 runs with your operating-system permissions; ordinary Python is not sandboxed.

Read Data & user control for the difference between local execution, cloud inference, and actions that affect external resources.