Open models
Built on open models, not one black box
Digby runs on leading open-weight models, picks the best one for each question, and cross-checks the answers that matter.
01
Custom-tuned open models
Digby starts from leading open-weight models and tunes them for accuracy, safety and the way Canadians work. Open weights mean Digby can run them on Canadian hardware — and you can run them on yours.
02
The right model per question
Writing, reasoning, code and research each favour different models. Digby routes every question to the one best suited to it, and a quick question doesn’t wait on a heavyweight model.
The lineup changes as better open models are released. Because they are open, adopting a new one is a configuration change — not a new contract.
03
A second opinion, built in
On contested or high-stakes questions an independent model checks the first answer before you rely on it. If it finds a real problem, the answer is updated and you can see what changed.
04
No lock-in
No single company can change prices, terms or availability underneath you. Self-hosted Digby works with any OpenAI-compatible model server.
Frequently asked questions
Which open models does Digby use?
Custom-tuned open-weight models from several leading labs, chosen per question. The lineup is updated as better open models are released.
Are open models as good as ChatGPT?
The best open-weight models are now competitive with leading closed models on many tasks. Combining several of them — and checking one with another — closes much of the remaining gap.
What does “open-weight” mean?
The model’s trained weights are published, so anyone with suitable hardware can run it — a Canadian data centre, a cloud provider or your own server.
Try asking Digby
Keep reading
Try Digby now.
Free, no account needed. Models and data stay in Canada.