Google Updates Gemma 4 Chat Templates with Tool Calling Fixes, Reduced "Laziness", and Flash Attention 4 for Hopper GPUs

Core summary

One sentence to understand this update

Google is enhancing Gemma 4's chat templates, introducing significant fixes for tool calling, mitigating "laziness", and enabling Flash Attention 4 support on Hopper GPUs.

Impact & opportunity

What this could mean

Developers using Gemma 4 can expect improved tool integration, more responsive model behavior, and better performance on Hopper GPUs.