Anyscale
The managed Ray platform for scaling AI and Python workloads
Overview
Anyscale is the managed platform from Ray's creators, running the proprietary Anyscale Runtime for distributed data processing, training, batch and online inference and LLM serving, hosted or in your own AWS, GCP, Azure, OCI, CoreWeave or Kubernetes account. $100 starting credit, then per-minute usage; Nscale acquisition pending.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Anyscale-managed infrastructure, limited regions
- CPU 0.0135, A100 4.96 Anyscale Credits/hr
- Business-hours support, 5 cases
- Runs in your VPC on VMs or Kubernetes
- Cloud marketplace invoicing
- 24x7 enterprise SLAs
- Volume discounts
- Use your own GPU reservations
What you can produce with Anyscale
- Managed Ray clusters with the Anyscale Runtime
- Distributed data processing and model training
- Batch and online inference and LLM serving
- Hosted or bring-your-own-cloud deployment
- Workspaces, autoscaling and lineage tracking
Anyscale is the natural choice for teams that already depend on Ray: it removes much of the operational work of running distributed AI workloads, adds a faster proprietary runtime, and can run inside your own cloud. Ray has a steep learning curve, which reviewers mention most, and newer GPUs are priced by quote. The pending acquisition by Nscale, a GPU cloud provider, may change its direction, though Anyscale says the brand will continue. Try it on the $100 credit if you need to scale Python beyond a single machine.
What we found
Anyscale is the company behind Ray, the open-source framework for scaling Python and AI workloads across many machines, which began at UC Berkeley's RISELab and which Anyscale donated to the PyTorch Foundation at the Linux Foundation in 2025. Ray is Apache 2.0-licensed, and Nscale's filings say it has been downloaded 740 million times. Anyscale's managed platform runs Ray for data processing, model training and fine-tuning, batch and online inference, LLM serving and agentic AI, using the Anyscale Runtime, its proprietary version of Ray formerly called RayTurbo, with workspaces that start in under a minute, autoscaling, lineage tracking and A/B rollouts. It can be hosted by Anyscale in limited regions or run in your own cloud account on AWS, Google Cloud, Azure, Oracle, CoreWeave or any Kubernetes environment, including on-premises.
Anyscale, Inc. was founded in 2019 in San Francisco, has a subsidiary in Bengaluru and is led by chief executive Keerti Melkote; its investors include Andreessen Horowitz, NEA, Addition and Intel Capital. On 28 July 2026 it agreed to a takeover by Nscale, the AI cloud provider, worth $1.65 billion in Nscale shares, with completion expected around Nscale's planned stock-market listing; Anyscale says it will keep its brand.
New accounts get $100 of credits. After that the hosted service bills per minute in Anyscale Credits per hour — 0.0135 for a CPU instance, about 0.57 for a T4 GPU, 1.36 for an A10G and 4.96 for an A100, with H100-class and newer GPUs by quote — and committed contracts bring volume discounts. Hosted support is business hours only; bring-your-own-cloud deployments get round-the-clock enterprise SLAs.
On AWS Marketplace it rates 4.5 from 24 ratings; of the ten reviews visible there, about six mention a steep learning curve and three mention cost. It suits teams already using Ray or needing to scale Python and AI workloads beyond one machine.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Anyscale's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Anyscale unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Anyscale directly →
Spotted something out of date? Suggest an update →
Alternatives to Anyscale
More in Code & Dev Tools
