BREAKING
Google Ships Gemma 4 Update
0
%
max prefill gain
0
%
faster TTFT
0
x
tokens per sec via MTP
Practical Fixes for Production
1
Tool calling accuracy
↓
2
Cleaner chat template
↓
3
Fix laziness
Vision Detail vs Efficiency
Default
280 tokens
●
Efficient processing
●
Lower cost
Max fidelity
1,120 tokens
●
Sharper OCR
●
Images up to 2.51MP
Reliability Is the New Battleground
Updated Weights Live on Hugging Face
AI NEWS BLITZ
Google is rolling out a major update to its open-weight Gemma 4 model family.