#diffusiongemma Tech News.

Google DeepMind released DiffusionGemma, a 26 billion-parameter open-weights model that uses diffusion techniques from image generators to produce text. The model generates entire paragraphs of tokens simultaneously rather than one token at a time. It runs on consumer hardware with just 18 GB of DRAM or VRAM and delivers up to 4x faster output performance.

Google released DiffusionGemma, an AI model that achieves 1,000 tokens per second. The model is available for free. This performance marks a significant speed milestone for text generation models.

Google released DiffusionGemma, an open weight Apache 2 licensed model (google/diffusiongemma-26B-A4B-it). NVIDIA hosts it for free on NIM cloud API. The model generated 2,409 tokens in 4.4 seconds, achieving at least 500 tokens per second.

Google DeepMind released DiffusionGemma, a new open model that generates text in parallel blocks rather than token by token. The 26-billion-parameter Mixture of Experts model activates only 3.8 billion parameters during inference. On an Nvidia H100, it produces over 1,000 tokens per second, roughly four times faster than similarly sized autoregressive Gemma models.

Google released DiffusionGemma, an open-weight text diffusion model with 26 billion parameters. It generates 256 tokens in parallel, running up to four times faster than traditional LLMs. On an H100, it exceeds 1,000 tokens per second. The mixture-of-experts architecture activates only 3.8 billion parameters per step.

Google released DiffusionGemma, a family of image generation models built on the Gemma framework. The models are 4x faster than other Gemma models. They are available on Hugging Face and Kaggle under a commercial-friendly license.

Google released DiffusionGemma, an experimental 26B-parameter open model using text diffusion for faster generation. The model delivers up to 4x faster inference on dedicated GPUs compared to autoregressive models, enabling speed-critical, interactive local workflows.