Google DeepMind has released DiffusionGemma, a 26B open model that uses text diffusion instead of autoregressive decodin...

Google DeepMind has released DiffusionGemma, a 26B open model that uses text diffusion instead of autoregressive decoding. The approach generates entire blocks of text in parallel, delivering up to 4x faster generation on GPUs than standard models. It fits in 18GB of VRAM and runs at 1000+ tokens per second on an H100. https://arstechnica.com/google/2026/06/googles-latest-diffusiongemma-open-ai-model-comes-with-a-4x-speed-boost/ #AIagent #AI #GenAI #AIResearch

Read Original

Related