gemma-4-e2b-it-qat-OptiQ-4bit

mlx-community/gemma-4-e2b-it-qat-OptiQ-4bit

Built with mlx-optiq, the MLX-native toolkit to quantize, fine-tune, and serve LLMs locally on Apple Silicon, no PyTorch and no cloud. Try the Lab · All OptiQ quants · Docs

A 4-bit mixed-precision MLX quant produced by mlx-optiq, built on Google's quantization-aware-trained (QAT) Gemma-4 base. OptiQ's sensitivity-guided per-layer bit allocation is applied on top of weights that were trained to survive low-bit quantization, and it still beats a uniform 4-bit quant of the same QAT base by +2.09 Capability Score points.

This is a quant of google/gemma-4-E2B-it-qat-q4_0-unquantized. Per-layer bit-widths come from a KL-divergence sensitivity pass on a six-domain calibration mix (prose, reasoning, code, agent, tool-call, constraint-bearing instructions). Sensitive layers go to 8-bit, robust ones stay at 4-bit.

Quantization details

Capability Score

Six-metric mean (MMLU, GSM8K, IFEval, BFCL, HumanEval, HashHop), scored against a uniform 4-bit quant of the same QAT base. That comparison isolates what the mixed-precision allocation adds, holding the base fixed.

OptiQ adds +2.09 points over uniform 4-bit on this QAT base, close to the +2.12 it adds on the vanilla gemma-4-e2b-it base, so the per-layer allocation keeps paying off even after QAT has made the weights more quantization-robust. The mixed quant is 5.24 bits-per-weight (about 4.9 GB on disk) versus 4.0 bits-per-weight (about 2.4 GB) for uniform 4-bit: the gain comes from spending the extra budget on the layers that need it.

Usage

mlx-lm loads it directly for text:

from mlx_lm import load, generate
model, tokenizer = load("mlx-community/gemma-4-e2b-it-qat-OptiQ-4bit")
print(generate(model, tokenizer, "Explain mixed-precision quantization.", max_tokens=256))

Image+text input and the speculative drafter run through mlx-optiq:

pip install mlx-optiq
optiq serve --model mlx-community/gemma-4-e2b-it-qat-OptiQ-4bit \
            --drafter google/gemma-4-E2B-it-qat-q4_0-unquantized-assistant

The same repo loads text-only under stock mlx-lm and image+text under optiq. The bf16 vision tower rides in optiq_vision.safetensors, which mlx-lm ignores (it globs model*.safetensors), so both paths work from one artifact.

Quantize your own

This quant was produced by mlx-optiq. Point it at any Hugging Face model to get the same sensitivity-aware mixed precision:

pip install mlx-optiq
optiq convert <hf-model-id> --target-bpw 5.0 --candidate-bits 4,8
optiq lab   # or open the full local workbench: chat, compare, quantize, fine-tune

License

Gemma Terms of Use. Built on google/gemma-4-E2B-it-qat-q4_0-unquantized.