Nemotron Labs Dory 1.0 (4B)

Nemotron Labs Dory 1.0 4B is NVIDIA's open-weight looped language model, a normalized GPT with interleaved sliding-window and full-attention layers whose shared recurrent block can be run multiple times at inference to trade compute for accuracy. It is text-in, text-out, English-first, supports up to 256K context and tool calling, and was pre-trained on 6 trillion tokens. Weights released 2026-10-16 under OpenMDW-1.1 in BF16, MXFP8, NVFP4 and GGUF formats.