Why Ollama
Ollama is Wilson's default AI backend. It downloads and runs open-weight language models on your hardware — no API keys, no cloud accounts, no per-token fees. Your financial data never leaves your machine.
Ollama supports Metal on Mac and CUDA on Linux/Windows, so inference is fast even on consumer hardware. Wilson auto-detects Ollama at startup and picks the best available model for each task.
The Ollama library has over 12,000 models. Wilson works best with small, fast models in the 2–7B parameter range — they handle categorization, anomaly detection, and structured output without needing a data-center GPU.
What you get
12,000+ open models available from the Ollama library
Metal and CUDA GPU acceleration for fast inference
Native tool calling — Wilson's agent loop works out of the box
Deploy fine-tuned models via custom Modelfiles
Getting started
$ curl -fsSL https://ollama.com/install.sh | sh
$ ollama pull qwen2.5:3b
$ wilson
> Ollama detected. Using qwen2.5:3b.
Recommended models
These models are tested with Wilson and work well for financial analysis tasks.
Model
Size
Best for
qwen2.5:3bllama3.2:3bmistral:7bgemma2:2bphi3:miniRun AI on your terms.
Install Ollama, pull a model, and Wilson handles the rest. No cloud account required.