Sovereign Inference DirectoryAbout made. ↗
← All guides

Buying guide

API access without reserved compute

How to separate usage pricing from minimum spends, platform fees and capacity commitments.

made. editorial · 2026-11-09 · 3 min read

Define your entry cost

The useful question is not just “What is the price per million tokens?” Ask what you must pay before your first production request, what you pay in a quiet month, and whether you are committed to capacity or a contract term.

Keep access and billing separate

Pay-per-token does not necessarily mean self-service. An option can require sales approval, a workspace subscription or a negotiated agreement. Record those requirements alongside usage rates.

Compare the same workload

Input and output token prices differ. Caching, batch processing, context length and tool use may change the bill. Compare a stated workload in the original billing currency; if showing AUD estimates, label the exchange rate and date.

What this directory prioritises

We prioritise shared, managed inference APIs without reserved capacity. Azure’s Standard deployment category and Bedrock on-demand invocation are research starting points. Exact model availability and Australian processing conditions still require separate verification.

Sources

Sources reviewed 09 Nov 2026. Product conditions may change.