BlinkDL announces RWKV-7 G1j, claiming gains in coding and agent tasks
BlinkDL describes the model as “100% RNN”—a recurrent neural network—and links to a demo at rwkv.com and model weights on Hugging Face.
TLDR
BlinkDL announced RWKV-7 G1j on September 1, claiming it is “much better” at agent, coding and STEM tasks, alongside broader improvements. The announcement describes it as “100% RNN” and shares a demo at rwkv.com and weights at huggingface.co/BlinkDL/rwkv7-g1.
Combined views
12K
2 Sources, first seen 30d ago
likes