Your Computer. Your Assistant. Meet Dojo’s Local Models.
No cloud model account? A capable computer can still be your starting point. Meet Dojo Tiny and Dojo Small: private, on-device assistants for everyday work in Dojo Solo.
A good computer should be more than a waiting room for someone else’s server.
Maybe you cannot access a cloud AI provider. Maybe a subscription is not practical, your connection is unreliable, or you simply want some work to stay on your own machine. Whatever the reason, getting started with an assistant should not depend on having another account ready.
Dojo Workspace includes an on-device path: Dojo Tiny and Dojo Small. Download a supported model, select it for Dojo Solo, and have a conversation powered by your computer.
Two sizes. One practical starting point.
Dojo Tiny is the lighter option. Its current download is approximately 3.2 GB. It is a starting point for everyday questions, explanations, drafting, and exploring ideas when you want a compact local assistant.
Dojo Small takes a step up, with a current download of approximately 4.5 GB. It offers another choice when you have more room and want to try the larger local option.
Those are download sizes—not promises about total memory use. Running a model needs additional working memory, and longer conversations can increase that demand. Response speed depends on your hardware, the model, and what else your computer is doing.
Useful work, without a cloud model account
Start with something concrete: “Explain this function in plain English.” “Help me outline this document.” “Turn these notes into a clearer checklist.” “Walk me through this error.”
Small local models are most useful when the task is focused and the source material is manageable. Give them clear context, review the answer, and break larger work into smaller steps.
They are not a claim that a few gigabytes can replace every frontier model. Complex engineering, difficult reasoning, and high-stakes decisions still demand judgment and verification. The point is a useful assistant you can access—not a contest to pretend every model is identical.
What you need
Dojo’s on-device AI requires a usable supported GPU. A powerful processor alone is not sufficient. Supported installations cover Apple Silicon Macs and supported Windows and Linux systems; operating-system and graphics-driver compatibility still matter.
Open Settings → On-device models to inspect availability and install the local chat option you want. With the model installed and available, choose it in the model picker for your Dojo Solo conversation. The initial download requires a connection; local model inference can then run offline.
Offline chat does not make every workspace feature offline. Web browsing, cloud providers, downloads, and external integrations still need their own connections and permissions.
Local means something specific
With an on-device model selected, the model’s inference runs on your computer rather than sending that conversation to a cloud model for processing.
That does not mean every action an assistant can take stays local. Opening a website or calling an external service is a separate operation. Keep the provider and the tools you choose aligned with the privacy requirements of your work.
For people who have the hardware but not ready access to a cloud LLM, this is the invitation: start where you are, with the computer you already have. Dojo Solo can be a place to learn, draft, ask better questions, and move the next useful task forward.