# areal-project/AReaL

The RL Bridge for LLM-based Agent Applications. Made Simple & Flexible.

Repository: https://github.com/areal-project/AReaL
Canonical: https://ross.abutalabs.com/products/areal
Homepage: https://areal-ai.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: llm, llm-reasoning, machine-learning-systems, mlsys, reinforcement-learning, rl, agent, llm-agent
Last push: 2026-08-26T23:14:54+00:00

## Health v2 (maintenance only)
Score: 87/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 99, longevity 39
- inputs: {"age_days": 555, "days_push": 7, "days_rel": 8, "gap_med": 19.5, "n_releases_24m": 23}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5696, forks 583 (observed 2026-08-28T04:09:28.022252+00:00)

## What it is
AReaL is a large-scale asynchronous reinforcement learning system that bridges foundation model training with agent-based applications, supporting algorithms like PPO, GRPO, and DAPO on backends such as Megatron, FSDP, and SGLang. Version 2.0 refactors it into a microservice architecture with independent training, inference, agent, and weight-update services for scalable agentic RL.

## Use cases
- train reasoning LLMs with reinforcement learning
- run agentic RL post-training for LLM agents
- fine-tune models with GRPO or PPO
- train agents with online RL against black-box APIs
- scale asynchronous RL training across GPU clusters
- post-train multi-turn tool-using agents with verifiable rewards

## When to choose
- you need scalable asynchronous RL training for large language models
- you want to train reasoning, coding, or search agents with RL
- you need to mix RL algorithms with popular training and inference backends
- you want online RL training against black-box agent applications

## When to avoid
- you only need simple supervised fine-tuning without RL
- you lack multi-GPU or cluster infrastructure
- you need a lightweight single-GPU hobby setup

## Facets
- artifact type: framework
- maturity: active
- function: llm-training, reinforcement-learning, agent-framework, machine-learning, gpu-computing, llm-inference
- domain: reinforcement-learning, large-language-models, machine-learning, deep-learning, gpu-computing
- platform: python, cloud
- tags: agentic-rl, asynchronous-training, grpo, ppo, rlhf, post-training, sglang, megatron, fsdp, mlsys, ai-agents, gpu, linux, docker

## Member repositories
- areal-project/AReaL (main) score 87

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:28.022252+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:53:37.870455+00:00, confidence not recorded.
  - readme: https://github.com/areal-project/AReaL (fetched 2026-08-28T04:09:28.022252+00:00, sha 295f5a498758)
  - homepage: https://areal-ai.io (fetched 2026-08-29T08:49:15.876614+00:00, sha d2f04e31a021)
  - site_page: https://areal-ai.io/docs (fetched 2026-08-29T08:49:15.885982+00:00, sha 2116a5f92b04)
  - site_page: https://areal-ai.io/docs/en/intro.html (fetched 2026-08-29T08:49:15.887887+00:00, sha d8edd5011ddb)
  - site_page: https://areal-ai.io/docs/en/version_history.html (fetched 2026-08-29T08:49:15.889565+00:00, sha 6a4bd8bd59e3)
  - registry_pypi: https://pypi.org/pypi/areal/json (fetched 2026-08-29T08:49:15.891569+00:00, sha ef574644acf0)
- Data as of 2026-08-30T08:39:29.467469+00:00.