Skip to content

RWKV v6 Finch 14B

An open RWKV model tuned for long-context workloads

Visit official site

What RWKV v6 Finch 14B is

The model is described as the sixth and largest RWKV release, built to handle long text more efficiently through data-dependent token shift and time mixing.

What you can do with it

  • Manage long-term memory while processing prompts
  • Generate text for multilingual chat
  • Serve translation and generation tasks

Getting started

  • Download weights and training code from the project page
  • Set up the required hardware and software
  • Test with the provided inference service

Pros & cons

✓ What we like

  • Long-term memory management is improved
  • GPU cluster time donations support further training

! What to watch out for

  • Results are reported against the Eagle 7B baseline
  • The model requires weights and code to be downloaded

FAQ

How large is the training set?

The material states a dataset of 1.42T tokens.

Who recognises the project?

The material names the Linux Foundation.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review