mlx-community/gemma-4-e4b-it-OptiQ-4bit

Point base_model at the upstream model and mark this repo as a quantization

Sync chat template from google/gemma-4-e4b-it (Google canonical, published 2026-07-09)

index: register the bf16 vision tower so stock VLM loaders (mlx-vlm, oMLX, LM Studio) can find it

Correct BFCL + Capability: the AST checker could not match array-typed arguments

Correct BFCL + Capability: the AST checker could not match array-typed arguments

Move OptiQ sidecars under optiq/ so *.safetensors loaders (mlx-vlm, LM Studio) skip them

fix: audio_tower conv weights to MLX channel-last layout (full-VLM loader compat)

docs: OptiQ brand spelling

card: remove em-dashes, ensure funnel

card: add mlx-optiq funnel banner + CTA

Restore vision_config + optiq_vision marker for image input

Add OptIQ vision sidecar (image+text support, v0.2.0)

Add kv_config.json (per-layer mixed-precision KV cache, 5.0 BPW target from OptiQ kv-cache sensitivity analysis). Requires optiq>=0.1.3 runtime for the RotatingQuantizedKVCache shim.

v0.1.0: 6-metric Capability Score + OptIQ vs U4 deltas

weights re-quant for mlx-lm 0.31 KV-shared Gemma-4 arch: optiq_metadata.json

weights re-quant for mlx-lm 0.31 KV-shared Gemma-4 arch: config.json

weights re-quant for mlx-lm 0.31 KV-shared Gemma-4 arch: model.safetensors.index.json

weights re-quant for mlx-lm 0.31 KV-shared Gemma-4 arch: model-00002-of-00002.safetensors

weights re-quant for mlx-lm 0.31 KV-shared Gemma-4 arch: model-00001-of-00002.safetensors

Fix chat template: emit multimodal placeholders in tool messages

card: link blog post for calibration mix instead of bare optiq.jsonl reference

card: drop loose IFEval row, keep strict only

v0.1.0 sweep: requant with 5.0-BPW ceiling, fresh benchmark numbers

Update Article link to https://x.com/latent_node/status/2028412948167942334

Normalize model card: remove twitter ref, lowercase 'optiq', add multimodal stripping note

Clean codelion/source refs from model card

v0.0.9: fix quant weight round-trip (eager-eval), route to gemma4_text, updated model card

Strip vision/audio metadata (text-only OptIQ variant cleanup)

Upload folder using huggingface_hub

initial commit