Ray Review, Pricing & Features

Ray is an open-source framework for scaling Python and AI workloads from a laptop to a cluster. See features, licensing, Anyscale pricing, and FAQs.

Category
AI Infrastructure & MLOps
Pricing
Free and open source (Apache 2.0 license); Anyscale sells a separate paid managed platform built on Ray, from Free (open source); Anyscale's managed platform pricing is usage-based and not publicly listed
Verified
Not yet
Last updated
July 19, 2026
Founded
2017
Headquarters
Ray originated at UC Berkeley's RISELab; Anyscale, the company founded by Ray's creators, is headquartered in San Francisco, California, United States
Free PlanAPIOpen SourceAISelf-Hosted

Overview

Ray is an open-source distributed computing framework, developed at UC Berkeley's RISELab, that lets Python and AI/ML workloads scale from a single laptop to a cluster spanning thousands of GPUs or CPUs. It is released under the Apache 2.0 license and is free to use and self-host.

The framework was commercialized starting in 2019 when its creators, Robert Nishihara, Philipp Moritz, and Ion Stoica, founded Anyscale, a San Francisco-based company that leads Ray's ongoing open-source development while also selling a separate managed cloud platform built on top of it.

Key Features

Ray Core provides distributed computing primitives, tasks, actors, and objects, that let developers parallelize Python code with minimal modification, backed by an automatically managed cluster architecture with a head node coordinating scheduling and worker nodes executing tasks.

On top of Ray Core, purpose-built AI libraries cover distributed data processing (Ray Data), model training (Ray Train), hyperparameter tuning (Ray Tune), model serving including LLM serving (Ray Serve), and reinforcement learning (RLlib), making Ray a single framework capable of handling most stages of a machine learning pipeline.

Pricing

The open-source Ray framework itself is completely free under the Apache 2.0 license, and organizations can self-host and scale Ray clusters on their own infrastructure or any major cloud provider at no licensing cost.

Anyscale, the company founded by Ray's original creators, sells a separate managed platform that adds enterprise governance, autoscaling, developer tooling, and support. Anyscale's pricing is usage-based and not published as a fixed rate card; prospective customers need to contact Anyscale directly for a quote.

Key Features

Pros & Cons

Pros

  • Free and open source under the permissive Apache 2.0 license
  • Scales seamlessly from a single laptop to clusters with thousands of nodes
  • Used in production by major AI companies, including reported use in ChatGPT-related infrastructure
  • Comprehensive library ecosystem covering data, training, tuning, and serving in one framework
  • Large, active open-source community with 1,000+ contributors and 34,800+ GitHub stars

Cons

  • Distributed systems concepts add a learning curve compared to single-machine Python tools
  • Self-managing Ray clusters without Anyscale requires nontrivial DevOps expertise
  • Debugging failures across a distributed cluster can be more difficult than single-node debugging
  • Enterprise-grade support and managed infrastructure are only available through the separately priced Anyscale platform

Pricing

Frequently Asked Questions

What is Ray?

Ray is an open-source framework for scaling Python and AI applications from a single laptop to a distributed cluster, developed originally at UC Berkeley's RISELab.

Is Ray free to use?

Yes, Ray is free and open source under the Apache 2.0 license. Anyscale sells a separate paid managed platform built on top of it.

Who created Ray?

Ray was created by researchers Robert Nishihara, Philipp Moritz, and Ion Stoica at UC Berkeley's RISELab; the same team founded Anyscale in 2019 to commercialize it.

What is Anyscale's relationship to Ray?

Anyscale is the company founded by Ray's creators. It leads Ray's open-source development and sells a separate managed enterprise platform built on the open-source framework.

What is Ray used for?

Ray is used for distributed model training, hyperparameter tuning, batch inference, model and LLM serving, reinforcement learning, and general distributed Python computation.

Is Ray connected to OpenAI's ChatGPT?

Public reporting has described Ray being used in infrastructure associated with training and scaling systems related to OpenAI's ChatGPT.

What programming languages does Ray support?

Ray is primarily Python-native; Ray Serve also offers a Java API for cross-language model serving.

How large is the Ray open-source community?

As reported on ray.io, the project has more than 34,800 GitHub stars, over 40,000 downloads, and 1,000-plus contributors.

Related Tools