AX-Qwen3.6-27B-MLX-AXQ-6bit-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 at BF16 in the checkpoint (or a bound sidecar when present).

AXQuant checkpoint Tier 1 certified — artifact edition v3. The exact v3 weight set passed measured size, matched-reference quality, zero-fallback conversion, and safe default-runtime gates on df-macbookpro-m5. Read the certificate and exact hashes.

MTP acceleration Tier 2 certified (scoped). On df-macbookpro-m5 with AX Engine 6.14.0, greedy MTP-off/on streams are identical and decode-heavy authorizing profiles clear ≥1.20× token-weighted and ≥1.10× prompt-median speedup (agent-coding 1.258× / 1.112×, long-form general 1.233× / 1.250×). Tier 2 certificate.

Product default remains direct fallback for the safe Tier 1 route. Short-answer chat is not an authorizing universal speed claim. Use the formal Qwen linear MTP exact / certification-candidate contract to exercise the certified acceleration path.

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. When that collapse happens, AutomatosX does not publish a separate misleading 4bit sibling for that base.

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-6bit-MTP \
  --revision v3 \
  --local-dir ./AX-Qwen3.6-27B-MLX-AXQ-6bit-MTP

Allow at least 20.73 GB of free disk space. The v3 tag identifies this certified artifact; pin the resolved Hub commit for production reproducibility. The previous development checkpoint remains available at v2.

Run with MLX-LM

python -m pip install -U mlx-lm
mlx_lm.generate \
  --model AutomatosX/AX-Qwen3.6-27B-MLX-AXQ-6bit-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.

Serve with AX Engine

After installing AX Engine, download the complete repository (see AXQuant for conversion, certificates, and model-card tooling) and serve the local directory:

ax-engine serve ./AX-Qwen3.6-27B-MLX-AXQ-6bit-MTP --port 31418

AX Engine is the authority for the AXQ runtime contract and native MTP sidecar. Tier 1 default-route smoke (AX Engine 6.13.5) keeps policy 4, MTP inactive, and direct fallback active. Scoped Tier 2 MTP acceleration is certified on AX Engine 6.14.0 under the formal Qwen linear MTP exact / certification-candidate contract (decode-heavy suites only). Product default remains direct fallback. Native model-manifest.json status: included.

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-6bit-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, dwq.
• Group sizes used by quantized assignments: 64.
• MTP sidecar: 15 tensors, 424.70M parameters, 0.85 GB, BF16.
• Vision sidecar: 333 tensors, 460.73M parameters, 0.92 GB, BF16.
• Vision weights: protected BF16 sidecar.
• 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 text generation 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.

• Quality evidence is relative to the matched uniform-6 reference on reproducible AXQuant suites; it is not a third-party benchmark, a universal quality guarantee, or a BF16-equivalence claim.

• “Stable default runtime” means the bound text path passed model loading, inference, and the fail-closed route gate on the certification machine. It is not a promise that every context, application, or third-party runtime is defect-free.

• MTP acceleration is opt-in under the formal Qwen linear MTP exact / certification-candidate contract (scoped Tier 2 on decode-heavy suites). Product default remains direct fallback. Short-answer chat is not a universal speed claim. Outside AX Engine, use a sidecar-aware runtime; oMLX/MTPLX compatibility does not extend the AX Engine certificate.

• Vision weights are 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.

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

Provenance and audit files

• Public Tier 1 certificate: verdict, thresholds, environment, and exact LFS hashes.

• 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 compatibility metadata; runtime checks remain separate evidence.

• axquant_mtp_sidecar_manifest.json: MTP tensor provenance.

• axquant_vision_sidecar_manifest.json: protected vision tensor provenance.

• model-manifest.json: AX Engine native tensor manifest.

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. If an OptiQ repository is published separately, it uses 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.