Run a model
Use the project's supported Ollama and local-model routes. Choose a model that fits your memory, storage and workload. Downloading a model and running it are separate from browsing this website.
Linux first · local AI · public source
OpenZero brings a model runtime, an operator panel and an OpenAI-compatible API together for builders who want their own AI setup. It is a self-hosted source project: you supply the computer, model and configuration.
Use the project's supported Ollama and local-model routes. Choose a model that fits your memory, storage and workload. Downloading a model and running it are separate from browsing this website.
The project documents a Super Panel for runtime health, model controls, configuration and local status. Follow the version-specific instructions for the installation you choose.
An OpenAI-compatible API gives applications and scripts a documented connection route. Configure the client's endpoint and the model actually available on your node.
Local inference uses memory, disk space and electricity. Larger models can need considerably more resources. Begin with a small task and a model that fits your machine.
A local model request can stay on your machine, while optional remote APIs and web tools use external services. Local installation alone does not make every workflow offline.
You operate the runtime, updates and network exposure. Keep the service on its intended interface and follow the project's release and security instructions.