// 1 transmissions tagged with #quantization-aware-training
Google unveiled Gemma 4 quantization‑aware training models that shrink size by up to 4× and keep accuracy within 1‑2 % of the full‑precision baseline, targeting smartphones and laptops.
