Open market for AI inference

The price floor
for AI inference.

One OpenAI-compatible endpoint. Every provider competes to serve each request and you pay the winning price. Want a specific one? Pin it. Got spare capacity? List it and earn. Sellers keep 95%.

OffersThe floor

The problem

Same model. Wildly different prices.

Most developers overpay without ever seeing it. They integrated one endpoint, shipped, and never looked back — while the price of the exact same model kept moving across providers.

01

Overpaying by default

The same open model is often priced several times apart across providers, and the gap moves week to week. Wired to one endpoint, you pay whatever it charges — on every single token.

02

Lock-in keeps it that way

One SDK, one bill, one rate limit. Switching providers means a rewrite, so you stay put even when a better route opens up.

03

One outage, no second route

When your only provider degrades or rate-limits you, your product goes down with it. There is nowhere to reroute.

04

Comparing is homework

Every provider ships a different API, pricing model, and latency profile. Working out who is best today is a spreadsheet — and it goes stale by Friday.

The solution

Every provider competes for your prompt

Floor is an open inference marketplace behind one endpoint. Every active offer for your model competes on every request — scored on five axes, highest score wins — and if the winner falters, the next best takes over. Your stack never changes.

One endpoint

Integrate once. Reach the whole market.

Point your existing OpenAI SDK at Floor and every seller on the market sits behind one key and one bill. No new client, no per-provider plumbing.

·
Providers behind it
·
Models served
·
Best-route savings
providers in one out

scored on five axes · the best route wins

Smart routing

Every competing offer is scored on five axes — price, latency, uptime, liquidity, and health. The highest score carries the request, and the full breakdown comes back with the response.

Automatic failover

If the winning offer errors or rate-limits, Floor reroutes to the next best one before a single byte streams. No retries in your code, no silent switches mid-stream.

Marketplace pricing

Sellers compete to serve each model, so you pay the winning offer's price — the market floor, no markup. Floor keeps a flat 5% of the seller's settlement; sellers keep 95%. Nothing is added to your bill.

OpenAI-compatible API

Change one line. Keep your code.

Point your existing OpenAI SDK at Floor. One base URL, one key, and every model on the market is within reach.

quickstart.pylive
from openai import OpenAI

client = OpenAI(
    base_url="https://api.floorcompute.com/v1",
    api_key="flr_live_••••",
)

# Same SDK. Floor finds the best route.
resp = client.chat.completions.create(
    model="llama-3.3-70b",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)
Exposed surface
  • /v1/chat/completions
  • /v1/completions
  • /v1/embeddings
  • /v1/models

Pricing

You pay the market floor. Nothing on top.

No subscriptions, no marked-up tokens. Providers compete for every request and you pay the winning offer's price. Floor keeps a flat 5% fee out of the seller's settlement — it's never added to your bill.

You pay
$0.00/Mtok

the winning offer’s price, per 1M tokens

Offer price (you pay)
$0.600
the market floor · routed now
=
Seller settlement
$0.570
95% to the provider
+
Marketplace fee
$0.030
flat 5% · our only revenue

Pay by card (Stripe)·USDC top-ups coming soon·Free flr_test_ sandbox keys

Roadmap

From one endpoint to the default routing layer

Built in the open, shipped in phases. Each one carries the network further toward being the default routing layer for open-model inference.

  1. Phase 01shipping

    One endpoint

    One OpenAI-compatible endpoint in front of every provider, backed by a live model catalog and routing you can see straight through.

    • OpenAI-compatible API
    • Model catalog
    • 5-axis routing
    • Console dashboard
  2. Phase 02shipping

    Routing that learns

    Every offer is scored on price, latency, uptime, liquidity, and health — the highest score wins. Failover reroutes around unhealthy offers before a stream starts, and every request comes with analytics.

    • 5-axis scoring
    • Pre-stream failover
    • Health scored 0 to 100
    • Routing analytics
  3. Phase 03shipping

    The marketplace is open

    The sell side is open to anyone. Sellers self-onboard, list competing offers from their own endpoints, and settle on every request they carry. They keep 95%.

    • Seller onboarding
    • Competing offers
    • Two-sided settlement
    • Seller payouts
  4. Phase 04

    The network layer

    The default routing layer for open-model inference: on-chain seller payout rails, multi-region scale, and decentralized reputation for every offer.

    • On-chain payouts
    • Multi-region scale
    • Decentralized reputation
Settled daily

Every request has a floor price. We find it.

Every provider wants your prompt. We make them compete for it — on price, latency, uptime, liquidity, and health. The best route wins, on every request.