AReaL logo
AReaL
★ 5.9k
vs
chidori logo
chidori
★ 1.4k

AReaL vs chidori

AReaL: AReaL is a fully asynchronous reinforcement learning system from Tsinghua IIIS and Ant Group for training large reasoning and agentic models. Version 2.0 splits it into training, inference, agent and weight-update microservices, and black-box agent apps can be trained online by pointing their base_url at AReaL.; chidori: Chidori runs agents written as plain async TypeScript inside one Rust binary with an embedded pure-Rust JavaScript engine. Every LLM call, tool call and HTTP request goes through the runtime as a recorded host call, so a run can be checkpointed, replayed byte-identically with zero LLM calls, paused for human input for minutes or days, and resumed in a new process after a crash. TypeScript and Python SDKs talk to it over HTTP.

01

TL;DR

AReaL logo
AReaLLLM Infra · ★ 5.9k

RL fine-tuning of reasoning models on math and coding tasks

chidori logo
chidoriLLM Infra · ★ 1.4k

Debugging non-deterministic agent bugs by replaying the exact run

02

Side-by-Side Comparison

Field
AReaL logoAReaL
chidori logochidori
Category
LLM Infra
LLM Infra
Stars
★ 5.9k
★ 1.4k
License
Apache-2.0
Apache-2.0
Install Command
git clone https://github.com/areal-project/AReaL && cd AReaL && pip install uv && uv sync --extra cuda
curl -fsSL https://raw.githubusercontent.com/ThousandBirdsInc/chidori/main/scripts/install.sh | sh
Updated
1d ago
6d ago
Open Source
Yes
Yes
Tags
agent, agent-framework, llm
durable-execution, agent-framework, typescript
03

Features

AReaL logoAReaL
01Fully asynchronous RL training, reported 2.77x faster than synchronous systems in v0.3
02Online RL for black-box agents by replacing base_url and api_key
03Microservice architecture in 2.0 with separate training, inference, agent and weight-update services
04SGLang or vLLM inference backends, Ray clusters and Ascend NPU support
05Examples for math, SWE, search agents, tau2-bench and OpenClaw agents
chidori logochidori
01Replay any recorded run byte-identically with zero LLM calls
02Checkpoints at every host safepoint, so a killed run resumes in a new process
03chidori.input() and named signals suspend a run to disk while waiting for humans
04Commit a recorded run as a zero-cost integration test
05One self-contained Rust binary: no Node, Deno or V8 required
04

Use Cases

AReaL logoAReaL
↳RL fine-tuning of reasoning models on math and coding tasks
↳Training multi-turn agents such as search or customer-service agents
↳Continuously improving an existing agent app via online RL
chidori logochidori
↳Debugging non-deterministic agent bugs by replaying the exact run
↳Long-running agents with approval steps that may wait for days
↳Regression-testing agent behavior in CI without paying for tokens
05

Best For

AReaL logoAReaL
EssentialTrending
chidori logochidori
—
FAQ

FAQ

What is the difference between AReaL and chidori?
Both AReaL and chidori are in the LLM Infra category. AReaL has 5.9k stars, while chidori has 1.4k stars.
Which is better, AReaL or chidori?
The best choice depends on your use case. Choose AReaL if RL fine-tuning of reasoning models on math and coding tasks, and chidori if Debugging non-deterministic agent bugs by replaying the exact run.
Is AReaL free or open source?
Yes, AReaL is open source on GitHub (Apache-2.0).
Is chidori free or open source?
Yes, chidori is open source on GitHub (Apache-2.0).
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
→

Related