Google Ships Gemma 4 Update Boosting Inference Speed and Fixing Tool Calling

Google has begun rolling out a substantial update to its open-weight Gemma 4 model family, delivering faster inference on NVIDIA Hopper GPUs and targeted fixes for tool calling, vision handling, and incomplete responses. The refresh, announced on July 15, 2026, folds in feedback and contributions gathered since the models first shipped in April, and the updated weights are available immediately through the Gemma 4 collection on Hugging Face.