- What changed
- Together AI released preemptible compute for Together GPU Clusters at 50% of the on-demand rate with sub-hourly billing and a five-minute termination grace period.
- Why you should care
- Preemptible GPU nodes reduce cost for interruption-tolerant batch jobs and short experiments.
- Your move
- Watch. Monitor workload stability and reclaim frequency before moving critical pipelines to preemptible nodes.
- What to watch next
- Updates on the general availability rollout or changes to node reclamation performance under high demand.
- Event
- release
- Event date
- Sep 10, 2026
- Relevant to
- General AI readers
