
Community collective · Hong Kong, Hong Kong
RWKV
An open-source community behind the RWKV recurrent language model architecture, hosted as a Linux Foundation AI & Data project.
Hong Kong, Hong Kong
- Type
- Community collective
- Based in
- Hong Kong, Hong Kong
- Key papers
- 1
- Featured projects
- 8
About
RWKV started as a project by independent researcher Bo Peng, who leads it, and grew into a community project under the Linux Foundation AI & Data.
The architecture has gone through versions named Dove, Eagle, Finch and Goose. RWKV-7 Goose, described in a March 2025 arXiv paper, generalizes the delta rule with vector-valued gating and supports state tracking. Models of 0.4B, 1.5B and 2.9B were released under Apache 2.0.
Contributors come from EleutherAI, Recursal AI and several universities.
Mission and ambitionsRWKV aims to offer an efficient alternative to transformers with linear-time inference.
Known for
- 01
RWKV architecture
- 02
RWKV-7 Goose
- 03
Linux Foundation project
- 04
Efficient RNN-style LLMs
Milestones
- 2023
RWKV-4
First publicly scaled RWKV models.
- 2023
Joins Linux Foundation
Project joined the Linux Foundation in September 2023.
- 2025
RWKV-7 Goose
Released March 2025 under Apache 2.0; the paper has 18 authors including Ruichong Zhang, Daniel Goldstein and Eric Alcaide.
Open source
CPU inference for RWKV.
Infinite context trainer.
Core RWKV repo.
ONNX converter.
- RWKV/v5-Eagle-7B-HFmodel
Eagle 7B.
Small Pile model.
RWKV-7 2.9B.
- RWKV/EagleX-WorldContinueddataset
Training data.
Research papers
Notable releases
Openness
Open weights and code (RWKV-LM, RWKV-Runner).
People
- Bo PengCreator and project lead
First author of the RWKV-7 paper.
Listed from the lab's own pages and public profiles.
Contact and community
rwkv.com links the community Discord and GitHub.
Main community Discord linked from rwkv.com.
Sources
Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.



