Who you are buying inference from
QuickSilver Pro is an OpenAI-compatible inference API for open-weight and frontier models, operated by MachineFi Inc. One key, one prepaid balance, one billing relationship, priced below the standard published per-token rate on most of the catalog.
The company
- Legal entity
- MachineFi Inc., a Delaware corporation
- Address
- 68 Willow Rd, Menlo Park, CA 94205, USA
- Payments
- Processed by Stripe. Charges appear on your statement as MACHINEFI INC.
- Infrastructure
- Dedicated bare-metal in Europe (OVH), with a US edge.
- Contact
- hello@quicksilverpro.io
What we are building
We are an inference-infrastructure company, not a reseller with a markup. Today the gap between our price and the published list rate comes from a narrow operational surface and tight engineering: a curated catalog, a small number of routes to keep healthy, and no cost we are not willing to explain.
The next step is our own serving layer on dedicated H100/H200 capacity. Open weights are public, which is precisely why they are the long game: we can run and optimise them ourselves, and that is where the next price cut comes from. Closed frontier models cannot be tuned that way, so we hold those below the vendor's own list price instead. Both live behind the same key.
The current state of that roadmap, including what is live and what is not, is on the status page. We would rather publish a plan that can be checked against reality than a promise that cannot.
How to check what we claim
Every price claim on this site names its reference rate and links to the page that publishes it. Everything else you might reasonably want to verify has a page:
- Trust & verifiability — our commitments against silent model substitution and quantization, and probes you can run yourself
- Pricing table — every row's reference list price, linked to its source
- pricing.json — the same catalog, machine-readable
- Status — live per-model uptime and latency
- Changelog — what shipped and when
- Terms · Privacy · DPA