BREAKING
Google releases Gemma 4 12B
0B
Parameters (Dense)
0K
Max context tokens
0GB
RAM/VRAM with quantization
Encoder-free design
1Text, image, audio in
2Linear projection
3Unified model
4Lower latency
Benchmark scores
GPQA Diamond78.8
LiveCode Bench72
BBEH53
Praise vs caveats
StrengthsPraise
Local coding agent on Mac
256K context in 16GB
Smooth video and audio
CaveatsConcerns
44 tokens/s on M5 Pro
Trails Qwen 3.6 family
Narrower knowledge breadth
Local AI fills the middle
AI NEWS BLITZ
Google DeepMind has launched Gemma 4 12B, an open multimodal model that runs on a laptop.