我们推出 Nemotron-Labs-Diffusion，这是一种三模式语言模型，在单一架构内统一了自回归、扩散和自推测解码。通过联合自回归-扩散目标进行训练，Nemotron-Labs-Diffusion 能够切换模式，在不同部署场景和并发水平下维持高吞吐量。我们的研究表明：(1) 自回归与扩散目标是互补的：扩散改进了前瞻规划能力，而自回归则提供了从左到右的语言先验。(2) 在自推测模式下，扩散负责草稿生成，自回归负责验证，在接收率和实际设备效率上均优于多 token 预测方法。(3) 光速分析进一步展示了扩散的长期潜力，在最优采样器下，每次前向传播可生成比自推测多 76.5% 的 token。我们的 Nemotron-Labs-Diffusion 系列模型规模扩展至 3B、8B 和 14B 参数，包括基础版、指令版和视觉语言版，在准确率和速度上均持续优于最先进的开源自回归和扩散语言模型。例如，Nemotron-Labs-Diffusion-8B 每次前向传播生成的 token 数量是 Qwen3-8B 的 6 倍，且准确率相当，在 GB200 GPU 上使用 SGLang 运行 SPEED-Bench 时，吞吐量提升了 4 倍。
We introduce Nemotron-Labs-Diffusion, a tri-mode language model (LM) that unifies AR, diffusion, and self-speculation decoding within a single architecture. Trained with a joint AR-diffusion objective, Nemotron-Labs-Diffusion can switch modes to sustain high throughput across deployment settings and concurrency levels. Our study shows that (1) AR and diffusion objectives are complementary: diffusion improves lookahead planning, while AR provides left-to-right linguistic priors. (2) In self-speculation mode, diffusion drafts while AR verifies, outperforming multi-token prediction (MTP) methods in both acceptance rate and real-device efficiency. (3) A speed-of-light analysis further demonstrates diffusion's long-term potential, with up to 76.5% more tokens per forward pass than self-speculation under an optimal sampler. Scaling to 3B, 8B, and 14B parameters, our Nemotron-Labs-Diffusion family, including base, instruct, and vision-language models, consistently outperforms state-of-the-art open-source AR and diffusion LMs in both accuracy and speed. For example, Nemotron-Labs-Diffusion-8B decodes 6x more tokens per forward than Qwen3-8B with comparable accuracy, translating to 4x higher throughput on SPEED-Bench with SGLang on a GB200 GPU.