Downloads · 30 days
31
17% of all-time downloads
xprtaker/gemma-4-E2B-unimodal-JANG-Q4-lite
gemma-4-E2B-unimodal-JANG-Q4-lite is a machine learning model from xprtaker. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as apache-2.0.
This model is a stripped, text-only (unimodal) configuration of Google's Gemma 4 E2B IT (Instruction-Tuned) model, quantized to JANG Q4 using vMLX Studio for optimized execution on Apple Silicon.
Downloads · 30 days
31
17% of all-time downloads
All-time downloads
186
Public
Parameters
4.6B
2.7 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors2.7 GB · 99%
How the weights are stored.
U324.6B · 100%
From the Hugging Face model README
This model is a stripped, text-only (unimodal) configuration of Google's Gemma 4 E2B IT (Instruction-Tuned) model, quantized to JANG Q4 using vMLX Studio for optimized execution on Apple Silicon.
By surgically removing the non-text modalities (vision and audio projection/embedding layers), we have significantly lowered the model's memory footprint and accelerated local text inference speed while retaining 100% of its base reasoning and instruction-following capabilities.
This model is designed for highly efficient, local text-only inference on Apple Silicon (macOS/iOS). It is ideal for local assistants, coding help, reasoning tasks, and text generation where vision and audio processing are not required.
This model will not work for any tasks requiring image, video, or audio inputs. Because the multimodal projection layers (like vision_projection and audio_projection) have been permanently stripped from the weight dictionary, sending media files to this model will result in errors.
xprtaker/gemma-4-E2B-unimodal-JANG-Q4-lite.To run the model directly from your terminal using the optimized MLX runtime:
# Install the runtime
pip install "jang[mlx]"
# Run inference
python -m jang.inference --model xprtaker/gemma-4-E2B-unimodal-JANG-Q4-lite