AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8 — 8.71 BPW measured main

An AXQuant (AXQ) mixed-precision MLX checkpoint for Apple Silicon, converted directly from the BF16 source model. The language path is quantized while the vision tower are preserved at BF16 in the checkpoint (or a bound sidecar when present).

Development evidence — not a certified AXQuant release. This package has conversion and artifact-integrity records, but it does not publish measured quality, long-context, kernel-speed, or MTP-speed evidence. Do not interpret the AXQ product label as a benchmark claim.

Model details

Property Value
Base model openbmb/MiniCPM-V-4.7-35B-A3B
Source revision 896c2d454672476a7b45dcc5e3f0c700a0218577
Product family minicpm-v-4.7
Source architecture MiniCPMV4_7ForConditionalGeneration (mixture of experts (MoE)); text path optimized
Main-model parameters 35.21B logical parameters
Quantizer AXQuant 2.3.0
Hub budget class MXFP8
AXQuant base precision class 9p4bpw
Planned storage-adjusted BPW 9.4198
Measured main-model BPW 8.7148
Measured total BPW 8.7148
Safetensors weight size 38.36 GB
Approximate complete download 38.49 GB
Configured maximum context 262,144 tokens; practical limits depend on unified memory
Primary MLX runtime MLX-VLM
AX Engine native execution Unsupported by the tested AX Engine 7.6.4 manifest generator
MTP present False
Vision present True
Audio present False

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

Choosing an AXQ pack

This family publishes MXFP4, MXFP8, and one CUDA NVFP4 W4A16 pack. The two MLX packs use the same role policy at different block formats. Protected tensors stay at higher precision, so measured BPW is the figure to compare. The NVFP4 format reference uses a CPU backend and is not a CUDA inference checkpoint validation.

Sibling Intended trade-off
MXFP4 sibling Same AXQ role policy at the other MLX block format. Vision, merger, and hybrid linear-attention tensors stay at source precision. No MTP tensors were detected.
CUDA NVFP4 W4A16 CPU reference backend encodes eligible 2D weights as NVFP4; fused 3D experts, vision, and protected tensors stay at source precision. CUDA runtime and quality are unverified.

See the AutomatosX collections for the family catalog, or the complete index.

Download

python -m pip install -U huggingface_hub
hf download AutomatosX/AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8 --local-dir ./AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8

Allow at least 38.49 GB of free disk space. Pin the resulting Hub commit in reproducible deployments rather than relying indefinitely on main.

Run with MLX-VLM

python -m pip install -U mlx-vlm
python -m mlx_vlm.generate \
  --model AutomatosX/AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8 \
  --image ./image.png \
  --prompt "Describe this image." \
  --max-tokens 128 \
  --temperature 0.0

The protected vision tower and AXQ language decoder are loaded together by MLX-VLM. The artifact records MLX 0.32.4; runtime QA is reported separately from model-quality claims.

AX Engine status

The tested AX Engine 7.6.4 manifest generator rejects the source model type minicpmv4_7. This package does not include model-manifest.json; AX Engine inference is not supported or validated for this artifact. The AX Engine fields in axquant_runtime.json record the attempted compatibility check, not a successful runtime check.

Quantization layout

Main-weight precision Parameters Share
8bit 33.10B 94.00%
bf16 2.11B 6.00%
  • Quantization methods: affine, bf16.
  • Group sizes used by quantized assignments: 32.
  • MTP sidecar: not included.
  • Vision sidecar: not included.
  • Vision weights: protected BF16 in main shards.
  • 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 231/231 recorded module conversions succeeded; 0 fallbacks
AX Engine native manifest not generated; tested AX Engine rejects minicpmv4_7
Quality versus BF16 or uniform baselines Not published; no quality-retention claim
MTP acceptance and speed not measured; no MTP speedup claim
AX Engine kernel evidence unmeasured
Vision-language quality Not evaluated or claimed; vision tensors are preserved at BF16
Speech-recognition quality Not applicable
Long-context quality 262,144-token capacity is config metadata, not a validated claim
Release certification Not certified; formal AXQuant M0-M8 gates are not closed

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.

  • 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.

  • The tested AX Engine manifest generator rejects minicpmv4_7; no AX Engine inference support is claimed.

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

Provenance and audit files

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 openbmb/MiniCPM-V-4.7-35B-A3B model card for license terms, model limitations, and responsible-use guidance.

Downloads last month
204
Safetensors
Model size
35B params
Tensor type
BF16
·
U32
·
MLX
Hardware compatibility
Log In to add your hardware

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for AutomatosX/AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8

Quantized
(78)
this model

Collections including AutomatosX/AX-MiniCPM-V-4.7-35B-A3B-MLX-AXQ-MXFP8