Google Releases Experimental AI Model DiffusionGemma
Google has released DiffusionGemma, an experimental AI model that generates text in blocks rather than word-by-word. The model is up to 4x faster than Gemma 4, but its output quality is inferior. It's open-sourced for developers and researchers, aiming to push speed and hardware efficiency.
Key points
- Google has released DiffusionGemma, an experimental AI model that generates text in blocks rather than word-by-word.
- The model is up to 4x faster than Gemma 4, with speeds of 1,000+ tokens per second on NVIDIA H100 and 700 on an RTX 5090.
- DiffusionGemma's output quality is inferior to Gemma 4, making it more of an experimental tool than a finished product.
- The model is open-sourced under the Apache 2.0 license, targeting developers and researchers rather than everyday users.
Google's latest AI model, DiffusionGemma, takes a different approach to text generation. Unlike most chatbots, which write one word after another, DiffusionGemma generates a whole block of text at once and then refines it until it becomes readable. This approach allows for significant speed increases, with DiffusionGemma reaching speeds of 1,000+ tokens per second on NVIDIA H100 and 700 on an RTX 5090.
While this is a notable achievement, the output quality of DiffusionGemma is still inferior to that of Gemma 4. As a result, the model is more suited for experimental purposes than everyday use. Google has open-sourced DiffusionGemma under the Apache 2.0 license, making it available to developers and researchers who can help push the boundaries of this new approach to text generation.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.