Slurm Instant Clusters Live
After months in beta and incredible feedback from our community, Slurm Instant Clusters are now fully live on Runpod. What used to take hours or days to set up now happens in seconds.
- Turnkey HPC Infrastructure. Launch production-ready clusters in just a few clicks.
- True Multi-Node Performance. Built for distributed training and large-scale simulations.
- Simple Web Management. Create, scale, and manage your entire cluster from one interface.
- Pay-As-You-Go. No idle cluster costs, just pure compute as you need it.
Model Store in Beta
We're launching Model Store in beta, a caching system that eliminates model download times when starting workers.
When deploying large models, you're stuck between embedding them in Docker images or downloading at startup. Both mean longer waits and charges during model loading.
Model Store places models on host machines before workers start. When you specify a model from Hugging Face, our scheduler prioritizes hosts with your model already cached. If cached, workers start instantly. If not, the system downloads before your worker starts billing.
New Public Endpoints
.webp)
.svg)