AX-Qwen3.6-35B-A3B-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 as BF16 sidecars when present.

AXQuant checkpoint Tier 1 certified on df-macbookpro-m5 for Hub commit 7b9ff47abfb8be01e636f516edb0226aa25ea1cc. Agent-coding and general quality retention both 1.000 vs the matched uniform-6 reference. Size vs uniform-6 is 0.928× (about 7% smaller). Certificate.

MTP acceleration Tier 2 is not certified. Formal greedy exactness is achievable with MoE MTP loaded, but decode-heavy speed gates (≥1.20× weighted / ≥1.10× prompt-median) are not met (agent ~1.09× / 0.88×; long ~1.01× / 1.01×). Product default remains direct fallback. Certification index.

Model details

Property	Value
Base model	Qwen/Qwen3.6-35B-A3B
Source revision	995ad96eacd98c81ed38be0c5b274b04031597b0
Product family	qwen3.6
Source architecture	Qwen3_5MoeForConditionalGeneration (mixture of experts (MoE)); text path optimized
Main-model parameters	35.11B logical parameters
Quantizer	AXQuant 1.2.0
Hub budget class	6bit
Artifact edition	v2
AXQuant base precision class	6bit
Planned storage-adjusted BPW	6.0000
Measured main-model BPW	5.7595
Measured total BPW, including MTP	6.0001
Safetensors weight size	26.96 GB
Approximate complete download	26.99 GB
Configured maximum context	262,144 tokens; practical limits depend on unified memory
MLX-LM compatibility	Standard text inference, compatibility level B
AX Engine native execution	Not established; no validated native manifest is included
MTP present	True
Vision sidecar present	True

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.

Sibling	Intended trade-off
4bit sibling	Lower-storage AXQ budget; check its exact BPW
6bit sibling	Higher average precision near the 6-BPW budget

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

Download

python -m pip install -U huggingface_hub
hf download AutomatosX/AX-Qwen3.6-35B-A3B-MLX-AXQ-6bit-MTP --local-dir ./AX-Qwen3.6-35B-A3B-MLX-AXQ-6bit-MTP

Allow at least 26.99 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-Qwen3.6-35B-A3B-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.

AX Engine status

AX Engine on df-macbookpro-m5 loads this checkpoint for Tier 1 certification. Product default remains direct fallback. Do not claim MTP acceleration for this MoE pack until Tier 2 speed gates pass. MoE MTP load requires an engine that accepts mlp.experts.gate_up_proj packing (see ax-engine PR for the load fix).

Quantization layout

Main-weight precision	Parameters	Share
4bit	24.70B	68.69%
6bit	8.75B	24.35%
8bit	701.90M	1.95%
bf16	1.80B	5.01%

• Quantization methods: affine, bf16.
• Group sizes used by quantized assignments: 32, 64.
• MTP sidecar: 19 tensors, 844.64M parameters, 1.69 GB, BF16.
• Vision sidecar: 333 tensors, 446.57M parameters, 0.89 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

Check	Status
Planning evidence	architecture_prior
Calibration	none; the allocation is based on architecture priors
Quantizer execution	469/469 recorded module conversions succeeded; 0 fallbacks
AX Engine native manifest	not included
Quality versus matched uniform baseline	Certified on host — see certificate
MTP acceptance and speed	Exactness ok in formal A/B; speed not certified (agent ~1.09× / 0.88×; long ~1.01× / 1.01×)
AX Engine kernel evidence	unmeasured
Vision-language quality	Not evaluated or claimed; vision tensors are preserved at BF16
Long-context quality	262,144-token capacity is config metadata, not a validated claim
Release certification	Checkpoint Tier 1 certified; MTP Tier 2 not certified

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.

Modality	Claim	Supported	Reason
Vision	present-not-certified	true	vision present sidecar=['vision.safetensors']; mlx-vlm smoke failed on df-macstudio-m2 (mlx-vlm expects vision_tower.*; sidecar/layout mismatch). Text Tier 1 unchanged. Evidence: docs/certifications/evidence/modality-recert-capability-gated/results/AX-Qwen3.6-35B-A3B-MLX-AXQ-6bit-MTP.json
Audio	not-applicable	false	audio not supported (no tower config and no sidecar weights)

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 may be ignored outside AX Engine and its speedup is unmeasured for this exact checkpoint.

• 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 execution is not established because this package has no validated native manifest.

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

Provenance and audit files

• 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/Qwen3.6-35B-A3B model card for license terms, model limitations, and responsible-use guidance.