Last released Oct 23, 2026
Byte-level recurrent language model with mixture-of-experts, ternary/FP8 quantization, and a factored optimizer, on PyTorch.