See our collection for all Gemma 4 sizes and variants.

Run Gemma 4 with Keras 3: JAX, PyTorch, or TensorFlow

zeromodels/gemma-4-12b

Pure-Keras 3 conversion of google/gemma-4-12B for zeromodels. One implementation runs unmodified on TensorFlow / Torch / JAX. This is the 12B variant, served here as image + audio + text -> text via Gemma4UnifiedProcessor; weights are stored in bfloat16.

For model details, license, and usage terms, see Google's model card.

Gemma 4 family

✨ Quick start

Text-only

import os
os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from zeromodels.models.gemma4_unified import Gemma4UnifiedTextGenerate, Gemma4UnifiedTokenizer

model = Gemma4UnifiedTextGenerate.from_weights("zeromodels/gemma-4-12b")
tokenizer = Gemma4UnifiedTokenizer.from_weights("zeromodels/gemma-4-12b")

inputs = tokenizer([{"role": "user", "content": "Hello, who are you?"}])
outputs = model.generate(**inputs, max_new_tokens=64)
print(tokenizer.decode(outputs[0]))

Image + audio + text

import os
os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from PIL import Image
from zeromodels.models.gemma4_unified import Gemma4UnifiedConditionalGenerate, Gemma4UnifiedProcessor

model = Gemma4UnifiedConditionalGenerate.from_weights("zeromodels/gemma-4-12b")
processor = Gemma4UnifiedProcessor.from_weights("zeromodels/gemma-4-12b")

inputs = processor(conversation=[
    {"role": "user", "content": [
        {"type": "image", "image": Image.open("cat.jpg")},
        {"type": "audio", "path": "clip.wav"},
        {"type": "text", "text": "Describe the image and what you hear."},
    ]}
])
outputs = model.generate(**inputs, max_new_tokens=64)
print(processor.decode(outputs[0]))

Load any Gemma 4 variant the same way with from_weights("zeromodels/<variant>"):

Tips

• Set KERAS_BACKEND before importing Keras / zeromodels.
• Loads in bfloat16 by default. Pass load_dtype="float32" for full precision, or quantization="int8" to shrink further.
• See the Gemma 4 docs.
• Community / upstream weights still work via the hf: prefix: Gemma4UnifiedConditionalGenerate.from_weights("hf:google/gemma-4-12B").

Special Thanks

A huge thank you to the Google Gemma authors for creating and releasing these models.

License: Apache 2.0.