Google's Latest DiffusionGemma Open AI Model Comes with a 4x Speed Boost
Google says the experimental model matches other Gemma systems and runs about 4 times faster, with weights available on Hugging Face.
- Google released DiffusionGemma, an experimental model that delivers faster text generation than previous Gemma versions through simultaneous token prediction.
- Nvidia and Google collaborated to ensure optimization across diverse hardware setups, including enterprise systems like the H100 and DGX Spark, plus quantized RTX GPUs with efficient HBM.
- Model weights are available for download from Hugging Face under the same Apache 2.0 license as other fourth-generation Gemma models, enabling broad developer access.
- Google recently implemented Multi-Token Prediction drafters to utilize idle compute cycles; diffusion, however, is even faster than the MTP versions of Gemma.
- Diffusion models offer efficient compute usage, but face drawbacks in text generation; because language is discrete, errors can render tokens meaningless and force users to restart.
22 Articles
22 Articles
Google's new open-weights model brings image-generation tricks to AI text generation
The boffins on Google’s DeepMind team unveiled an experimental new language model this week that uses techniques originally developed for AI image generators to boost text output performance by as much as 4x when running on resource-constrained consumer hardware. It's free to download and you can run it with just 18 GB of DRAM or VRAM. The model, codenamed DiffusionGemma, is the latest addition to Google’s open weights model family. But unlike G…
Google's DiffusionGemma runs text 4x faster
GenAI image generators like Stable Diffusion do not draw a picture pixel by pixel from left to right. They start with noise and iteratively refine the entire image in parallel until it converges, in a process known as diffusion. For years, applying that same principle to text generation had remained out of reach at scale.Standard language models work like a typewriter: one token at a time, left to right, with no ability to revise a committed out…
Google's latest DiffusionGemma open AI model comes with a 4x speed boost
Another day, another AI model from Google. This time, Google DeepMind has released a new member of the Gemma 4 open model family, but it's fundamentally different from the rest of the lineup. DiffusionGemma doesn't generate outputs linearly like most AI models. Instead, it can produce an entire block of text in parallel. Google says this makes it faster and more efficient when running on local hardware like an Nvidia DGX or a humble gaming GPU. …
Google’s New Open Model Is a Win for Consumers and Proof Liberal Democracies Can Reclaim Open-Source AI
Google’s New Open Model Is a Win for Consumers and Proof Liberal Democracies Can Reclaim Open-Source AI DiffusionGemma shows American firms can still out-innovate China’s open-weight champions, if policymakers let builders build WASHINGTON, D.C. – Google this week released DiffusionGemma, an experimental open model published under a permissive Apache 2.0 license that generates text up to 4x… Source
Google has released an experimental open-source model, DiffusionGemma, which radically changes the traditional approach to text generation. Unlike standard models like Gemma 4, which write strictly sequentially—word by word—the new model generates an entire text array at once as a random set of "noisy" tokens, and then, in several passes, cleans and edits it until it is readable. Essentially, while conventional AI models write text sequentially,…
Google has presented without too much noise its new AI DiffusionGemma, an open model that changes the way to generate text to prioritize speed in local GPU, even accepting a loss of quality compared to Gemma 4. The movement of the great G fits better within the pulse of open models that are pushing from China, with Qwen or DeepSeek as unavoidable references, rather than within a direct comparison with GPT or Claude, where the battle is fought wi…
Coverage Details
Bias Distribution
- 100% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
