NVIDIA 发布 Nemotron-Labs-Sigma:3B/8B 连续扩散语言模型
NVIDIA AI Releases Nemotron-Labs-Sigma, a Continuous Diffusion Language Model with 14 checkpoints at 3B/8B
NVIDIA 把扩散式文本生成做成了能用的开源模型,3B/8B 共 14 个检查点随便下。解码比自回归快到 4.20 倍,做推理加速、本地部署的值得实测对比一下。
NVIDIA 发布 Nemotron-Labs-Sigma,一次性放出 14 个开源检查点,规格为 3B 和 8B 两种。这是语言建模路线上的新尝试:不同于 GPT 类模型逐 token 生成,Sigma 采用连续扩散方式生成文本。
技术上,Sigma 在一个学习得到的 16 维嵌入空间中对文本做去噪,并通过蒸馏得到少步采样器,解码速度最高达到自回归解码的 4.20 倍。这意味着在同等硬件下可以更快拿到生成结果,尤其适合对吞吐要求高的场景。
扩散语言模型此前多见于学术研究,NVIDIA 以开源检查点的形式将其工程化,给了开发者一个可与自回归开源模型对比的新选项,3B/8B 的规模也便于在消费级硬件上实测。
NVIDIA AI Releases Nemotron-Labs-Sigma, a Continuous Diffusion Language Model with 14 checkpoints at 3B/8B
NVIDIA released Nemotron-Labs-Sigma, 14 open 3B and 8B continuous diffusion language model checkpoints. Sigma denoises text in a learned 16-D embedding space, distills to few-step samplers, and decodes up to 4.20x faster than autoregressive decoding. The post NVIDIA AI Releases Nemotron-Labs-Sigma, a Continuous Diffusion Language Model with 14 checkpoints at 3B/8B appeared first on MarkTechPost .
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include NVIDIA, Nemotron-Labs-Sigma, diffusion models, open-source models.
Primary source verification: published directly by marktechpost with first-party authority. Cross-source corroboration: 11 independent sources have covered this development (including NVIDIA AI, Decoder, Jim Fan), confirming industry-wide signal strength.