AX-Qwen3.6-27B-MLX-AXQ-4bit-MTP

An AXQuant (AXQ) mixed-precision MLX checkpoint for Apple Silicon, converted directly from the BF16 source model. The language path is quantized while the multi-token-prediction (MTP) head and vision tower are preserved as BF16 sidecars when present.

AXQuant checkpoint Tier 1 certified on df-macbookpro-m5 for Hub commit f44a9eeebec0c488d0f42201c8763db770a1c0a8. Product class is 5p6bpw (mixed AXQ; marketing name 4-bit). Size vs uniform-4 is 1.207× under the class 1.25 budget; agent-coding quality retention 0.993, general 1.000. Certificate.

MTP acceleration Tier 2 certified (scoped). On df-macbookpro-m5 with AX Engine 6.14.0, greedy MTP-off/on streams match and decode-heavy profiles clear ≥1.20× / ≥1.10× (agent-coding 1.301× / 1.104×, long-form general 1.223× / 1.249×). Tier 2 certificate.

Product default remains direct fallback. Short-answer chat is not a universal speed claim. Formal route: Qwen linear MTP exact + certification-candidate opt-in.

Model details

This repository contains MLX Safetensors. It does not contain PyTorch or GGUF weights.

Choosing an AXQ pack

AXQ names describe a storage-budget product class, not one uniform precision applied to every tensor. Protected tensors remain at higher precision, so the exact measured BPW is authoritative. In particular, a 6bit-named mixed plan may retain 4bit as its base precision while selecting 6-bit, 8-bit, or BF16 for other tensors to meet an approximately 6-BPW total budget. Protection floors can also raise a 4bit-named pack close to (or above) a 6bit budget on small or heavily protected models.

See the AutomatosX MLX model catalog for related MLX and OptiQ alternatives.

Download

python -m pip install -U huggingface_hub
hf download AutomatosX/AX-Qwen3.6-27B-MLX-AXQ-4bit-MTP --local-dir ./AX-Qwen3.6-27B-MLX-AXQ-4bit-MTP

Allow at least 19.40 GB of free disk space. Pin the resulting Hub commit in reproducible deployments rather than relying indefinitely on main.

Run with MLX-LM

python -m pip install -U mlx-lm
mlx_lm.generate \
  --model AutomatosX/AX-Qwen3.6-27B-MLX-AXQ-4bit-MTP \
  --prompt "Explain mixed-precision quantization in three sentences." \
  --max-tokens 128 \
  --temp 0.0

MLX-LM compatibility covers standard text/backbone inference. It may ignore AXQuant runtime metadata and optional sidecars (vision.safetensors, mtp.safetensors); this command therefore does not establish MTP acceleration or vision-language quality. The artifact records MLX 0.32.0 and MLX-LM 0.31.3 from conversion.

AX Engine status

AX Engine 6.14.0 on df-macbookpro-m5 loads this checkpoint for formal certification. Product default remains direct fallback (safe Tier 1). Scoped Tier 2 MTP requires the formal Qwen linear MTP exact / certification-candidate contract (see certificate). MLX-LM remains the standard text inference path and does not by itself establish MTP acceleration.

Use the packaged Qwen MTP head with oMLX or MTPLX

Download the complete repository to a writable local directory. In oMLX 0.6.3rc2 or newer, add that directory, open Model Settings, choose Import MTP side-car, and then enable Lightning MTP. The import changes only the local copy so the sidecar tensors become visible through the checkpoint index. This is a text-path compatibility result; it does not certify VLM loading or vision quality.

MTPLX can consume the packaged sidecar directly:

mtplx quickstart \
  --model ./AX-Qwen3.6-27B-MLX-AXQ-4bit-MTP \
  --profile stable \
  --depth 1 \
  --reasoning off

mtplx_runtime.json declares the canonical qwen3-next-mtp execution contract. This enables strict runtime discovery; it does not establish oMLX or MTPLX exactness or speed certification.

Quantization layout

• Quantization methods: affine, bf16.
• Group sizes used by quantized assignments: 32, 64.
• MTP sidecar: 15 tensors, 424.70M parameters, 0.85 GB, BF16.
• Vision sidecar: 333 tensors, 460.73M parameters, 0.92 GB, BF16.
• Optimization scope: text-path.
• Support tier: convertible.

BF16 sidecars, when present, are included in total download size. Their presence does not by itself establish MTP acceleration or vision-language quality.

Evidence and validation status

Modalities (capability-gated)

Text checkpoint Tier 1 does not imply vision or audio quality. Vision present=true on a pack is not a quality pass.

Intended use and limitations

• Intended for local development and evaluation on Apple Silicon with MLX-compatible runtimes.

• No minimum unified-memory figure is claimed; loadability depends on model size, context length, KV-cache policy, runtime buffers, and other processes using unified memory.

• Architecture-prior allocation is not measured sensitivity. It must not be presented as measured model quality.

• MTP acceleration is opt-in under the formal exact contract; default remains direct fallback.

• Vision weights are byte-preserved at BF16, but this release does not claim validated VLM quality.

• The configured context window can require substantially more memory as the KV cache grows.

• AX Engine Tier 1 default is direct fallback; Tier 2 is scoped formal-route only.

• Upstream capabilities, limitations, biases, and responsible-use guidance still apply.

Provenance and audit files

• Public Tier 1 certificate: verdict, thresholds, environment, and digests.

• Public Tier 2 MTP acceleration certificate: decode-heavy exactness and speedup gates, formal env, comparison digests.

• axquant_manifest.json: package identity, byte accounting, runtime contract, software versions, and file checksums.

• axquant_plan.json: per-tensor precision decisions and planning evidence.

• axquant_quantizer_execution.json: conversion coverage and fallback records.

• axquant_runtime.json: declared AX Engine and MLX-LM compatibility metadata; runtime checks remain separate evidence.

• axquant_mtp_sidecar_manifest.json: MTP tensor provenance.

• axquant_vision_sidecar_manifest.json: protected vision tensor provenance.

All published provenance uses repository-relative paths. Local source paths are stripped before publication. The checkpoint was converted from BF16 rather than re-quantized from an OptiQ artifact. Parallel OptiQ repositories use a different quantizer and should not be assumed to have identical BPW or quality.

License

The checkpoint follows the upstream model license where applicable (often Apache License 2.0). See the Qwen/Qwen3.6-27B model card for license terms, model limitations, and responsible-use guidance.