LLM providers and BYOK

Pick a provider per role per request, and bring your own API key.

Choosing your LLM provider

On every run you pick the LLM provider for two roles:

• Worker — the model that drafts the formulation. This is where most of the heavy lifting happens. Stronger reasoning models tend to produce better first-draft formulations.
• Supervisor — the model that reviews the worker’s output and self-evaluates the final answer. A different provider here gives you a useful cross-check.

Feasible supports Anthropic, OpenAI, Google, and DeepSeek today, and is designed so additional providers can plug in without changes to the workflow.

Try a problem with two different worker providers and compare formulations side by side — they often differ in subtle ways (which auxiliary variables they introduce, how they encode logical constraints), and the differences are instructive. There is no “right” provider — pick what works for your problem.

BYOK — bring your own key

You can store your own API key for any of the supported providers under Settings → API Keys in the frontend, or via the API. Keys are encrypted at rest and only used for your requests. You retain full control:

• Use your own key for one provider, the Feasible-managed key (where available) for another.
• Rotate or revoke at any time.
• See, per provider, whether you have a key on file.

This is the same mechanism the team uses internally — there is no separate “managed-only” mode. If your organization has a contract with one LLM vendor, you can use Feasible end-to-end on that vendor.