Together AI Launches Preemptible GPU Compute at 50% of On-Demand Price

Together AI has launched a public preview of preemptible compute for Together GPU Clusters on Kubernetes in all regions, billed sub-hourly at a flat 50% of the on-demand rate.

Preemptible nodes can be reclaimed when capacity is needed elsewhere, with a five-minute drain window for checkpointing before removal.

The rate stays fixed rather than tracking a spot market, and the cluster automatically refills its preemptible target as capacity becomes available.