Gemma-4 — Vector Quantized (MLX) - a TheDrainFlorist Collection

TheDrainFlorist 's Collections

DeepSeek-V4-Flash — Vector Quantized (MLX)

GLM-5.3-Flash-VQ — Vector Quantized (MLX)

Qwen3.8-Flash-Next — Vector Quantized (MLX)

Qwen3.8-27B — Vector Quantized (MLX)

Gemma-4 — Vector Quantized (MLX)

Qwen3.6-35B-A3B — Vector Quantized (MLX)

Qwen3.5-397B-A17B — Vector Quantized (MLX)

Gemma-4 — Vector Quantized (MLX)

updated Sep 3

Data-free VQ builds of gemma-4 for Apple Silicon: the 26B MoE and the e4b with VQ per-layer embeddings. Stock mlx-lm. Built with VQLab.

TheDrainFlorist/gemma-4-26b-a4b-it-VQ-6.2bpw

Image-Text-to-Text • 7B • Updated Oct 11 • 115

Note 18.7 GiB, 6.2 bpw. Perplexity indistinguishable from bf16 on our referee.
TheDrainFlorist/gemma-4-e4b-it-VQ-PLE

Image-Text-to-Text • 6B • Updated Oct 11 • 77

Note 7.4 GiB. e4b with the per-layer embedding table vector-quantized (PLE).