Nemotron-Labs-Sigma: NVIDIA's open diffusion language model that writes text in blocks
NVIDIA's 8B continuous diffusion language model denoises text in parallel token blocks instead of generating one token at a time, with open weights under the NVIDIA Nemotron Open Model License.
Why it matters
First large-scale continuous diffusion LM at 3B/8B scale: block-parallel decoding in a 16-D embedding space, plus a tri-mode design that also runs autoregressive or self-speculative by switching the attention pattern.
What you could build with it
A research team could fine-tune the 8B base model into a fast draft generator for speculative decoding, pairing its parallel block output with a larger autoregressive verifier and measuring end-to-end latency on their own GPU fleet.
Does it hold up?
Too early to judge: the base model needs a CUDA GPU (CPU inference is unsupported) and the benchmarks are NVIDIA's own; the SFT checkpoint and plain-PyTorch loading make it tryable, but no independent speed claims exist yet.
Built with Nemotron-Labs-Sigma
- NVIDIA releases Nemotron-Labs-Sigma, a model that writes text in blocksalextech · Coverage of the 8B and 3B release: block-parallel denoising, DDPM/DDIM samplers, and the tri-mode design that also supports autoregressive decoding.
- SGLang serving guide for Nemotron-Labs-Diffusion on DGX Sparkgithub · The project repo's step-by-step guide to serving the 8B model on SGLang with linear self-speculation, plus a one-script benchmark evaluator.
Learn more
First spotted on article: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Edge0-145B: open 145B MoE that runs on a single consumer GPUEdge0 open-sourced 145B and 120B latent-MoE models derived from Qwen3.8-Flash-Next, plus an inference framework that…models · JEV 0.74Strata: run a 125B model on a gaming PCFree, MIT-licensed local inference engine that runs Qwen3.8-Flash-Next (125B MoE, about 6B active per token) on…models · JEV 0.73Google AI Edge Gallery: offline speech recognition and synthesis on-deviceGoogle added fully offline speech recognition and text-to-speech to AI Edge Gallery, with ML Kit APIs on Android and…models · JEV 0.73Soup: fine-tune an 8B LLM on a 4 GB laptop GPU from one YAML fileAn open-source CLI that fine-tunes LLMs from a single YAML file, with layer streaming that trains an 8B model on a 4…models · JEV 0.73Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.