Your models can deliver near-frontier performance at a fraction of the cost — but discovery is hard. Keld puts your capacity in front of live demand from enterprises and runs jobs on the best-fit model at the best price. Keld Trade is the trading platform you use to manage your orders and micro-batch matched work onto your fleet.
Keld Flow is a managed service — a capability we run on your behalf, not a separate product. You expose your model to Keld; Flow turns incoming demand into paced micro-batches that hit your fleet efficiently. You get a production-grade batch endpoint without writing a line of batching infrastructure, and you monetise idle capacity that would otherwise sit unused.
Keld provides the infrastructure for any AI model provider to offer a batch endpoint, without having to build it themselves.
Point Keld at your model. No batching code required.
Incoming demand is queued, batched and paced onto your fleet.
A production batch endpoint that monetises spare throughput.
The trading platform — place, cancel and reprice resting asks with algorithmic yield, plus a micro-batching layer that smooths matched jobs onto your fleet.
Keld Trade →Keld runs jobs from enterprises building apps and agents on the best-fit model at the best price — your capacity is in front of it.
Sell idle compute into live demand instead of letting it sit — a channel alongside your existing ones.
Keld surfaces live demand signals so you can allocate compute before spikes arrive. Track month-to-date revenue against forecast and catch utilisation shifts before they become missed opportunity.
We're onboarding AI Model Providers through the private beta. See Keld Trade, read the documentation, or get in touch.