Google's new open model DiffusionGemma generates text from noise instead of word by word
Quick Answer
Google has launched DiffusionGemma, a 26-billion-parameter model that generates text through diffusion rather than token-by-token, achieving around 1,000 tokens per second on a single H100 GPU.
Quick Take
While it operates four times faster than traditional autoregressive models, the output quality is lower, making it an experimental tool for developers.
Key Points
- DiffusionGemma uses a novel diffusion approach for text generation.
- The model has 26 billion parameters and generates text at 1,000 tokens per second.
- It is four times faster than comparable autoregressive models.
- Output quality is lower, limiting its current use to experimental purposes.
- Google targets developers with this new tool for further exploration.
Source Excerpt
Google released DiffusionGemma, a 26-billion-parameter model that generates text not token by token but through diffusion, similar to how image AI turns noise into a picture. According to Nvidia, it hits about 1,000 tokens per second on a single H100 GPU, roughly four times faster than comparable autoregressive models. The speed comes at a cost, though. Output quality is lower, so Google is positioning it as an experimental tool for developers for now.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
