Anyscale is a managed platform for distributed AI training and data processing, built on Ray. Usage-based pricing with a $100 free credit to start.
Anyscale is a managed platform for building and scaling data-intensive AI workloads, built on Ray, the open-source distributed computing framework. The company was founded in 2019 by Ray's original creators — Ion Stoica, Michael I. Jordan, Philipp Moritz, and Robert Nishihara — and is headquartered in San Francisco, California.
The platform orchestrates distributed model training across GPU clusters with elastic scaling, and provides pipelines for preparing multimodal data (video, image, text, audio) and generating embeddings at scale for search and retrieval. It supports multi-cloud orchestration across AWS, GCP, and Azure, and pools GPU resources dynamically across teams, with SSO, SAML, SCIM, and audit logging for enterprise governance.
Anyscale uses usage-based, pay-as-you-go pricing with no fixed monthly fee, charging for compute consumed — for example, roughly $0.0135/hour for CPU-only instances up to several dollars per hour for high-end GPUs like the A100. Volume-discounted committed contracts are available for larger commitments, and new customers receive $100 in credits. Deployment is offered either as a limited-region Hosted service or as Bring-Your-Own-Cloud (BYOC) with 24/7 enterprise SLAs across any cloud, region, or on-prem environment.
Anyscale is built on Ray, the open-source distributed computing framework created by Anyscale's founders.
Anyscale uses usage-based, pay-as-you-go pricing for compute, with volume-discounted committed contracts also available. New users get $100 in credits.
Anyscale was founded in 2019 by Ion Stoica, Michael I. Jordan, Philipp Moritz, and Robert Nishihara.
Yes, its Bring-Your-Own-Cloud (BYOC) option supports deployment on any cloud, region, or on-prem environment.