About Us
(the version that matches reality)
What we are
- A real-time router that hunts the lowest-cost GPU seats we can rent, then relays your request to whichever provider is cheapest at that moment.
- Up-front about the trade-offs: FP8 by default, zero certifications, zero SLAs, and some upstream providers may log or train on your data.
- One-click OpenAI-compatible endpoint that works as soon as you top up your account with USD.
- Extremely fast about adding new frontier open source models. Usually same day.
What we are not
- SOC 2, GDPR, HIPAA, or ISO-anything.
- A walled garden that logs everything you send. (We ourselves log only your raw tokens usage count, but we can't speak for every provider we route to.)
- A forever-stable service—prices, routes, and even upstream providers can change overnight.
How to use us responsibly
- 1.Great for hobby projects, MVPs, internal tools, and any workload that can tolerate occasional hiccups.
- 2.Not recommended for regulated data or customer-facing production where uptime guarantees matter.
- 3.If you need a locked-down provider or full-precision FP16, flip the "pin-provider" header or look elsewhere.
Roadmap (rough order)
- Reserved-capacity lanes for heavy users (no SLA, just better odds).
- Batch/async endpoints for even deeper discounts.
- Maybe—big maybe—an opt-in "enterprise tier" once we can afford audits.
Talk to the humans who run the boxes
- Email: Contact us