我们推出 DiffusionGemma，这是一款实验性的开放权重语言模型，利用离散扩散技术以极高速度生成文本。DiffusionGemma 并非逐 token 解码，而是并行迭代精炼 256 个 token 的块，从而绕开传统自回归（AR）大语言模型的顺序解码瓶颈。
我们并非从零训练，而是通过对 Gemma 4 混合专家模型进行微调得到 DiffusionGemma，该模型激活参数为 3.8B，总参数为 25.2B。我们计算高效的两阶段训练流程所用训练 token 预算不足原始 AR 模型总预算的 10%。
第一阶段使用监督微调来教授双向去噪，第二阶段则将强化学习与采样器蒸馏相结合，共同提升生成质量与推理效率。DiffusionGemma 在生成速度与模型能力之间的权衡上确立了新的帕累托前沿。在我们的完整评测套件上取平均，它每次前向传播可生成约 20 个 token，在单块 NVIDIA H100 GPU 上每秒可输出约 1,500 个 token，即便与采用最先进推测解码的 AR 模型相比也明显更快。
DiffusionGemma 还保留了原始模型对思考模式、多模态输入和长上下文的支持。尽管经过扩散微调，它仍能以仅轻微的性能下降进行 AR 生成，这为混合扩散-AR 解码指明了一条路径。
We introduce DiffusionGemma, an experimental open-weight language model that uses discrete diffusion to generate text at exceptionally high speed. Rather than decoding one token at a time, DiffusionGemma iteratively refines blocks of 256 tokens in parallel, avoiding the sequential decoding bottleneck of conventional autoregressive (AR) large language models. Instead of training from scratch, we obtain DiffusionGemma by fine-tuning the mixture-of-experts Gemma 4 model with 3.8B activated and 25.2B total parameters. Our compute-efficient two-stage training pipeline uses fewer than 10% of the starting AR model's total training token budget.
The first stage uses supervised fine-tuning to teach bidirectional denoising, while the second stage combines reinforcement learning with sampler distillation to jointly improve generation quality and inference efficiency. DiffusionGemma establishes a new Pareto frontier for the trade-off between generation speed and model capability. Averaged across our full evaluation suite, it generates around 20 tokens per forward pass and achieves roughly 1,500 output tokens per second on a single NVIDIA H100 GPU, which is substantially faster than AR models even with state-of-the-art speculative decoding.
DiffusionGemma also retains the starting model's support for thinking mode, multimodal inputs, and long contexts. Despite diffusion fine-tuning, it remains capable of AR generation with only minor performance degradation, suggesting a path toward hybrid diffusion-AR decoding.