Gemma 3
Google DeepMind — Gemma
partial detail 15 method fields published
Evaluation
Impact of image resolution ablation
effect of vision encoder input resolution (higher-resolution encoders use average pooling to reduce output to 256 tokens, e.g., 4x4 average pooling for the 896-resolution encoder) on a short-schedule 2B Gemma model over a few evaluation benchmarks
How it was run
Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.