Practical overview
When it fits
Use Skypilot when AI workloads span multiple infrastructure environments or when a team needs shared scheduling, scaling and orchestration for GPU, TPU or CPU capacity.
Who it is for
- AI teams running training, serving or other compute-intensive workloads.
- Infrastructure teams managing shared GPU, TPU and CPU capacity across clusters and clouds.
Documented uses
- Run distributed training and other machine-learning workloads on available infrastructure.
- Queue, schedule and recover multiple AI jobs.
- Serve large language models across managed AI infrastructure.
- Coordinate shared clusters and improve utilization through scheduling, binpacking and idle-resource cleanup.
- Provision compute across supported clouds, Kubernetes and Slurm through one interface.
Requirements & limits
The repository is public and licensed under Apache-2.0. Installation is documented through uv, pip, nightly builds or source. The documented interfaces include CLI, self-hosted and server-oriented infrastructure use.
- Supported environments include Kubernetes, Slurm and a broad list of cloud and infrastructure providers; provider choice determines the relevant integration.
- The tool is designed for infrastructure orchestration rather than a standalone model or end-user AI application.
- Existing GPU, TPU and CPU workloads are documented as supported without code changes.
- Scheduling features include gang scheduling, multi-cluster operation, binpacking, smart failover and automatic stopping of idle resources.
- Teams needing shared operation can use the documented deployment and resource-sharing capabilities.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 2 October 2026
- Original website title
- skypilot-org/skypilot · Official homepage title · retained in its source language

