RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose".

It's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctxlen, and free text embedding. And it's 100% attention-free, and a Linux Foundation AI project.