Google DeepMind releases DiffusionGemma, a 26B diffusion model 4x faster than autoregressive generation
DiffusionGemma uses parallel text diffusion instead of sequential token generation, achieving 1000+ tokens/sec on H100 GPUs with trade-offs in output quality.
Google DeepMind's Gemma 4 12B Brings Encoder-Free Multimodal AI to Consumer Laptops
Google DeepMind releases Gemma 4 12B, a 12-billion-parameter model with unified vision and audio processing that runs on 16GB consumer hardware.
Google's Agentic April: Cloud Next '26, Gemma 4, and a Two-Pronged AI Strategy
Google unveiled the Gemini Enterprise Agent Platform, eighth-generation TPUs, and open model Gemma 4 in a month-spanning push to dominate the agentic AI era.