AXIOM serves our own models and routed frontier models behind one OpenAI-compatible API. The free floor stays free. Frontier inference costs money to run, so it is metered — at the cost of the model plus a small margin that funds the work, falling as shared capacity grows.
| What | Price | Notes |
|---|---|---|
| Our own models (AXIOM-native) | free | Hosted, no account needed for the open rail. |
| The mesh, wiki & hive knowledge, the store | free | Shared commons. No paywall, no embargo. |
| Self-hosting the whole stack | free | Source and binaries are open. Run it yourself. |
Priced per million tokens, passed through at cost plus a small margin. The router records the actual upstream cost of every request; nothing is marked up in secret.
| Direction | Price / 1M tokens |
|---|---|
| Prompt (input) | €5.00 |
| Completion (output) | €15.00 |
Rates apply across the routed fleet; the cheapest capable model is chosen first and the decision is visible per request. Model-specific pricing is shown live in the console.
Prepaid, cancel anytime. A plan is just committed credit — you spend it at the metered rates above, with a little more back than you put in.
It covers the metered cost of the models, the servers, and the time to build the open stack — nothing else. No ads, no attention harvest, no data sale, no lock-in. Prices fall as shared capacity grows; the ratchet is applied to access itself. See the commitments in the manifesto.