Devtoolwatch BoardRay review
Ray is an open source AI compute engine with a core distributed runtime and libraries for scaling Python and ML workloads.
TL;DR
Ray is a framework for scaling Python and AI applications from a laptop to a cluster. It is for Python developers running compute-intensive ML work. Ray pairs a core runtime of tasks, actors and objects with libraries for data, training, tuning, reinforcement learning and serving.
- 44.2kGitHub stars
- +61new stars, week of Oct 25
- 3.7Mdownloads a week, +5% vs 4-week avg
- #25in Agent frameworks this week
Ray at a glance
- Category
- Agent frameworks
- Maintained by
- ray-project (organization, @ray-project on GitHub)
- License
- Apache 2.0
- Language
- Python, C++, Java
- Repository
- github.com/ray-project/ray
- Website
- ray.io
- PyPI package
- ray 2.59.0
- Latest release
- ray-2.57.1 on GitHub, Oct 21, 2026
- First published
- Oct 25, 2016
What Ray does
- Ray Core provides tasks, actors and objects as primitives for building and scaling distributed Python applications.
- Ray Data provides scalable datasets for ML, including processing of images, video, audio and other unstructured data.
- Ray Train runs distributed training, and Ray Tune runs scalable hyperparameter tuning.
- RLlib runs scalable, distributed reinforcement learning workloads.
- Ray Serve deploys models and business logic with independent scaling and fractional resources.
- Ray Dashboard monitors Ray apps and clusters, and the Ray Distributed Debugger debugs Ray apps.
How it works
Ray is installed as a Python package. Your Python code uses Ray Core primitives: stateless tasks, stateful actors, and immutable objects shared across a cluster. The same code scales from a laptop to a cluster. Ray runs on any machine, cluster, cloud provider, and Kubernetes. It can use mixed CPUs and GPUs in one pipeline.
How to install and start using Ray
With pip
pip install rayRequirements
- Python 3.10 or later
First steps
- Install Ray with pip. For nightly wheels, see the Installation page in the Ray docs.
- Read the Ray Core walkthrough to learn tasks, actors and objects.
- Choose a Ray AI library that fits your workload: Data, Train, Tune, RLlib or Serve.
- Monitor your app and cluster with the Ray Dashboard.
Who Ray is for
- Python developers who need to scale existing code from a laptop to a cluster
- ML teams running distributed training, hyperparameter tuning or reinforcement learning
- Teams serving models, including LLMs, or running batch inference on mixed CPUs and GPUs
Limitations to know
- Ray targets Python applications. The README says Python applications can be scaled with it.
- Ray is a general compute engine, not an agent-specific framework. Agent features are not described in the sources.
Is Ray growing? Usage data
In the week of Oct 25–31, 2026, Ray ranked #262 on the Devtoolwatch Board and #25 in Agent frameworks, with a momentum score of 36 out of 100. Its main package had 3.7M downloads that week, +5% against its own 4-week average.
Show the weekly numbers
| Week of | New stars |
|---|---|
| Aug 9, 2026 | 62 |
| Aug 16, 2026 | 67 |
| Aug 23, 2026 | 76 |
| Aug 30, 2026 | 74 |
| Sep 6, 2026 | 79 |
| Sep 13, 2026 | 92 |
| Sep 20, 2026 | 65 |
| Sep 27, 2026 | 51 |
| Oct 4, 2026 | 59 |
| Oct 11, 2026 | 63 |
| Oct 18, 2026 | 75 |
| Oct 25, 2026 | 61 |
On Hacker News
- Ray: Fast and Simple Distributed Computing – Build any application at any scale, 34 points and 0 comments (Jan 25, 2021)
The Board ranks tools by four signals each week: new stars, new stars for the repo's size, added downloads and download growth beyond the market. How the ranking works.
Ray alternatives
Other AI agent frameworks on the Board in the week of Oct 25–31, 2026, by momentum:
| Rank | Tool | Stars | Downloads/wk |
|---|---|---|---|
| #1 | Harness²Harness² is a Python framework from Google Research for recursive agent harnessing, refining an agent's prompts, tools, memory and… | 21.7k | – |
| #2 | HyperFramesHyperFrames is an open-source framework that turns HTML, CSS, media, and seekable animations into deterministic MP4 videos, built for AI… | 76.6k | 1.5M |
| #3 | AgentScopeAgentScope is a Python agent framework with an SDK for building ReAct agents and an agent service backend with a web UI. | 35k | 51.4k |
| #4 | AG-UIAG-UI is an open, event-based protocol that standardizes how AI agents connect to user-facing frontend applications. | 16.9k | 4.6M |
| #5 | LangGraphLangGraph is a low-level orchestration framework and runtime for building, managing, and deploying long-running, stateful agents in… | 44.2k | 18.5M |
| #6 | BifrostOpen-source, high-performance AI gateway that works as your LLM gateway, MCP gateway and agent gateway. One OpenAI-compatible API for… | 9.3k | – |
Recent Ray releases
- ray-2.57.1, Oct 21, 2026
- ray-2.58.1, Oct 21, 2026
- ray-2.59.0, Oct 2, 2026
- ray-2.58.0, Aug 23, 2026
- ray-2.57.0, Aug 11, 2026
Ray FAQ
Is Ray free and open source?
Yes. Ray's source code is on GitHub (ray-project/ray) under the Apache 2.0 license, so you can use it for free.
How do I install Ray?
Run pip install ray. The ray package needs Python 3.10 or newer.
Is Ray still maintained?
Yes. The ray-project/ray repository was last updated on Oct 21, 2026, and the latest release, ray-2.57.1 on GitHub, came out on Oct 21, 2026.
How popular is Ray?
Ray has 44.2k GitHub stars as of Nov 1, 2026. It gained 61 new stars in the week of Oct 25–31, 2026, against a 12-week average of 69 a week. Its main package had 3.7M downloads that week, +5% against its own 4-week average. That made it #25 among AI agent frameworks on the Devtoolwatch Board.
What are the best alternatives to Ray?
Other AI agent frameworks on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: Harness² (#1), HyperFrames (#2), AgentScope (#3), AG-UI (#4), LangGraph (#5). The alternatives table above has their numbers.
What is Ray used for?
Ray is used to scale Python and AI applications from a laptop to a cluster. Its libraries cover data processing, distributed training, hyperparameter tuning, reinforcement learning and model serving, including LLM inference.
What are the core abstractions in Ray?
Ray Core has three abstractions. Tasks are stateless functions executed in the cluster. Actors are stateful worker processes. Objects are immutable values accessible across the cluster.
Where can Ray run?
Ray runs on any machine, cluster, cloud provider, and Kubernetes. The homepage says it can scale from a laptop to thousands of GPUs and use mixed GPUs and CPUs.
Does Ray support LLM serving?
Yes. Ray Serve LLM and Ray Data LLM became generally available in Ray 2.59.0, with an upgrade to vLLM 0.27.0. Release 2.58.0 added KV cache and token aware request routing in Ray Serve LLM.
How do I debug a Ray application?
Ray provides the Ray Distributed Debugger for debugging Ray apps. It also provides the Ray Dashboard for monitoring apps and clusters.
Sources
Ray README badge
Maintainers can show the tool's weekly Board rank in their README:
[](https://index.devtoolwatch.com/tools/ray-project-ray/)How this page is made: Devtoolwatch writes each review from the project's README, releases, package metadata and website, with an AI model, then checks every claim against the source it cites and drops what the sources don't support. The numbers come from Devtoolwatch's daily GitHub snapshots, GitHub's stargazer history, and the npm and PyPI download APIs. Spotted an error? Email [email protected].