Pricing

Open-weights you run for free, plus a paid hosted API for teams that don’t want to run hardware. The model is free; the convenience and the private flywheel are what you pay for.

Open / Self-host

$0forever

Apache-2.0 open weights. Run them on your own hardware in one command.

✓Apache-2.0 weights — yours to keep
✓Run with llama.cpp, vLLM, or any OpenAI-compatible client
✓Your data stays on your machine unless you opt into training
✓A lapsed plan never bricks your local model
✓All niche models included

Get the install command

Hosted API

From $49per month · billed monthly

Managed inference — no hardware to run. OpenAI-compatible API, keys minted in seconds. Plans scale with the number of AI employees you run.

✓Starter — 1 AI employee · $49/mo
✓Pro — 3 AI employees · $199/mo
✓Scale — 10 AI employees · $699/mo
✓Agency — 30 AI employees · $1,999/mo
✓OpenAI-compatible /v1 · keys minted in seconds · no hardware

Private model — trained on your business

Customsetup + monthly hosting

Model-Training-as-a-Service: we train a private, single-tenant model on YOUR business — warm-started from the niche closest to your trade — and host it behind your API key. Never pooled with anyone else.

✓Private single-tenant model — your data only, never shared
✓Warm-started from your niche, then trained on your specifics
✓Call it as model: "private" with your existing key
✓Keeps improving from your consented usage (the real flywheel)
✓DPA + de-identification; deletion → retrain + version expiry

Train your model

Positioning: open-weights you can run anywhere, plus a paid hosted API and private models for teams that want managed inference and a flywheel that compounds on their own consented data.

Self-host is and will remain $0. Hosted plans are billed monthly via Stripe. Each plan includes generous monthly usage (Starter ~5M, Pro ~25M, Scale ~100M, Agency ~400M tokens); beyond that, overage is billed at $0.30 per million. Need a private single-tenant model or white-label? Talk to us.