Run open models on your own machine.
Ollama runs the models, Big-AGI is the workspace. Free, no API key, no lock-in. Point Big-AGI at your local Ollama and use open-weight models right next to Claude, GPT, and Gemini.
Free · Open source · Runs offline
Why Ollama in Big-AGI
Ollama, with a real workspace on top.
Private by default
Your prompts and replies stay on the machine that runs the model. Nothing leaves your network, nothing is logged at our edge.
No token bills
You bought the hardware once. Local inference is electricity, not a subscription, so you run as much as your machine can handle.
Hybrid on tap
Run local for the everyday turns, then reach for Claude, GPT, or Gemini in the same chat when you want frontier capability.
Beam across both
Beam runs your local model against frontier APIs in parallel, then merges the answers. Consensus is signal, disagreement is where you dig deeper.
Set it up
Three steps to your first local chat.
01
Install Ollama
Grab the installer for macOS, Windows, or Linux from ollama.com, then pull a model with ollama pull llama3.2. Your pick: Llama, Mistral, Qwen, Gemma, and more.
02
Open Big-AGI
Use the hosted workspace or self-host it. No account needed. Big-AGI becomes your one chat surface for local and cloud models alike.
03
Point it at Ollama
In Models, add Ollama and set the host to http://localhost:11434. Big-AGI auto-loads whatever you have installed. Add cloud keys later for hybrid runs.
Run models locally in Big-AGI.
Free and open source. Your models on your hardware, frontier models one click away, all in one workspace.
BIG-AGI
Resources
© 2026 Token Fabrics·Built with passion in San Diego