# Google DeepMind 发布 DiffusionGemma：无需从头训练即可构建文本扩散模型 - 来源：The Decoder：AI News（RSS） - 作者：Jonathan Kemper - 发布时间：2026-08-09 18:01 - AIHOT 分数：60 - AIHOT 链接：https://aihot.virxact.com/items/cmslnb7td03w5roqy97l7c4sc - 原文链接：https://the-decoder.com/googles-diffusiongemma-proves-you-dont-need-to-train-from-scratch-to-build-a-text-diffusion-model ## AI 摘要 Google DeepMind 发布 DiffusionGemma，将现有 Gemma-4-26B-A4B 改造为文本扩散模型，训练 token 预算不到原来的 10%。 ## 正文 Instead of training a new model from scratch, Google DeepMind retrofitted Gemma 4 into a diffusion model. The newly published report explains how it works and where the tradeoffs are. Google DeepMind released DiffusionGemma as a model in mid-June and has now followed up with the technical report. Unlike standard language models that generate text one token at a time, DiffusionGemma refines blocks of 256 tokens in parallel, similar to how image AIs pull a picture out of noise. On an Nvidia H100 accelerator, the model hits about 1,500 tokens per second. Building a new model from scratch wasn't necessary. The team started with the existing Gemma-4-26B-A4B and converted it into a diffusion model using less than ten percent of the original training token budget, according to the report. DiffusionGemma delivers several times the output speed of the Gemma 4 models and previous diffusion models while maintaining comparable accuracy. | Image: Google Two training stages balance quality and speed In the first of two steps, the model learns to reconstruct noisy text blocks from example data. A combined phase of reinforcement learning and sampler distillation follows, which Google calls SD·RL. Reinforcement learning typically boosts answer quality, while sampler distillation lets the model get by with fewer compute steps. Google merges both into a single process. Google DeepMind doesn't train DiffusionGemma from scratch but converts the finished Gemma 4 model through two training stages. | Image: Google According to the report, this combined approach raises quality on reasoning benchmarks by an average of ten points while nearly quadrupling the number of tokens per compute step. As a side effect, DiffusionGemma's answers run about 50 percent shorter, which further boosts speed. Bidirectional reasoning lets the model correct itself Standard language models have to commit to the first digit of an answer before they've worked through the reasoning. In a math problem from the report, Gemma 4 starts its response with "-1," realizes during its derivation that "-25" is correct, and tacks on a correction afterward. DiffusionGemma develops the answer and reasoning in parallel, so it can fix mistakes before the output is finalized. Unlike an autoregressive model, DiffusionGemma can correct an early wrong answer during later denoising steps. | Image: Google Sudoku solving works on the same principle, since every entry depends on entries that come later. After minimal fine-tuning, DiffusionGemma solves close to 85 percent of puzzles correctly, while the base model fails at the task entirely. Structured outputs like JSON or code repairs finish after just two to three refinement steps, according to the report, because the input already determines most tokens. DiffusionGemma also keeps its original ability to generate text word by word, letting users switch between both modes depending on the task. DiffusionGemma trails the autoregressive Gemma 4 on quality benchmarks but leads on output speed at about 1,500 tokens per second. | Image: Google Reasoning gaps and multi-user limits persist Absolute performance falls short of the autoregressive base model. Google points to several reasons for this. DiffusionGemma wasn't trained as a diffusion model from the start but was retrofitted after the fact. The subsequent training phase was relatively short, and the second step, SD·RL, prioritized speed over peak quality. The architecture, training data, and other settings were also carried over from the original Gemma 4 model, which aren't necessarily ideal for diffusion. The model occasionally gets stuck in repetition loops, producing individual words multiple times in a row. This is an artifact of the aggressively reduced compute steps. On multimodal tasks, DiffusionGemma sometimes forgets to close its reasoning section properly, which artificially drags down benchmark scores. The speed advantage also holds mainly for single-user scenarios. Once about 32 concurrent requests hit the model, standard language models catch up on throughput. Google explicitly calls DiffusionGemma an experimental model and says the release is meant to speed up research on text diffusion while giving the community a foundation for specialized, resource-efficient adaptations. The model is already being used by the startup Interfaze for multilingual speech recognition and in a research project on interactive radiology report generation. Google previously made the model available under an Apache 2.0 license on Hugging Face. Its predecessor is Gemini Diffusion, which Google demoed in May 2025. Read on for the full picture. Subscribe for hype-free coverage. Access to all THE DECODER articles. Read without distractions – no Google ads. Access to comments and community discussions. Weekly AI newsletter. 6 times a year: “AI Radar” – deep dives on key AI topics. Up to 25 % off on KI Pro online events. Access to our full ten-year archive. Get the latest AI news from The Decoder.