Lemura AI Labs
Models · 46Qwen3.5-122B-A10B-Abliterated-MLX-3

Qwen3.5-122B-A10B-Abliterated-MLX-3

image-text-to-text122BMLX3-bitapache-2.0
Download on Hugging Face
Downloads
2.4k
This month
198
Base model
Chompa1422/Qwen3.5-122B-A10B-abliterated
Updated
1mo ago

Lemura Labs

Qwen3.5-122B-A10B-Abliterated-MLX-3

Format Task Params Type License

Qwen3.5-122B-A10B-Abliterated-MLX-3 is an Apple Silicon-oriented MLX quantization of the abliterated Qwen3.5 122B-A10B vision-language model.

Publisher: Lemura Labs Hugging Face organization: Lemura Labs

This release is the 3-bit / group-size-32 build selected from our vision-quantized candidate run. This model quantizes both the language side and eligible multimodal/vision modules.

Source And Credits

This quant was produced from the local full-precision abliterated model derived from:

Thank you to the Qwen team for the base model, to Chompa1422 for publishing the abliterated source model that made this quantization work possible, and to Pliny the Liberator for the broader abliterated-model research culture that inspired this release.

Quantization

FieldValue
Runtime formatMLX / MLX-VLM safetensors
Quantization methodMLZTQ 0.2 custom MLX-VLM affine quantization
Weight bits3
Group size32
Modeaffine
Quant profilemlztq-0.2-visionq_w3-g32
Quant predicatevisionq_w3
Language weightsQuantized
Vision/multimodal weightsQuantized where MLX shape constraints allow
KV cacheTurboQuant runtime contract recorded; model weights are MLX affine quantized

Tensor spot checks from the selected artifact:

TensordtypeShapeMeaning
language_model.model.layers.0.linear_attn.in_proj_qkv.weightuint32(12288, 288)Language side quantized
language_model.model.layers.0.linear_attn.in_proj_qkv.scalesbfloat16(12288, 96)Language affine scales
vision_tower.blocks.0.attn.qkv.weightuint32(3456, 108)Vision side quantized
vision_tower.blocks.0.attn.qkv.scalesbfloat16(3456, 36)Vision affine scales

File Details

ItemValue
Safetensor shards12
Indexed tensor payload61,488,610,272 bytes
Local disk footprint57 GiB
Decimal payload size61.49 GB
Model indexmodel.safetensors.index.json
Quant manifestmlztq_manifest.json
Processor filesprocessor_config.json, preprocessor_config.json, video_preprocessor_config.json
Tokenizer filestokenizer.json, tokenizer_config.json, vocab.json, chat_template.jinja

Benchmarks

Benchmarks were run with the local MLZ deterministic VLM retention gate on 2026-05-02. Raw JSON, CSV, and Markdown benchmark artifacts are included under benchmarks/results/.

BucketCorrectTotalAccuracy
Quant health44100.0%
Text / reasoning66100.0%
OCR / document66100.0%
Vision general51050.0%
Overall gate212680.8%

Benchmark coverage:

BucketSources
Quant healthSafetensor load, vision path availability, text canary, vision canary
Text / reasoningMMLU-Pro, MMLU, GSM8K
OCR / documentChartQA, DocVQA, local PDF pages with images/text
Vision generalRealWorldQA, AI2D, MMMU Accounting, MMMU Biology, MMMU Physics

Candidate comparison from the same run:

CandidateWeight bits / groupPayload GBQuant healthText / reasoningVision generalOCR / documentOverallStatus
visionq-w6g326 / 32107.40100.0%83.3%50.0%100.0%76.9%kept locally
visionq-w5g325 / 32deleted100.0%100.0%40.0%83.3%73.1%pruned
visionq-w4g324 / 3276.79100.0%83.3%40.0%100.0%73.1%kept locally
visionq-w3g323 / 3261.49100.0%100.0%50.0%100.0%80.8%selected
visionq-w2g322 / 32deleted100.0%16.7%20.0%50.0%38.5%pruned

Inference Engines And Apps

This is an MLX / MLX-VLM safetensors repository for Apple Silicon. It is not a GGUF, AWQ, GPTQ, EXL2, or Transformers fp16 repository.

Engine / appStatus for this repositoryNotes
MLX-VLMIntended reference pathBest target for image-text inference because this artifact keeps the VLM processor files and MLX-VLM tensor layout. Requires a loader/runtime version with Qwen3.5 MoE VLM support and affine quantized multimodal weights.
LM Studio on Apple SiliconIntended app target, runtime-dependentLM Studio's unified MLX engine uses mlx-lm for text generation and mlx-vlm for vision embeddings. Use a recent LM Studio build with MLX support; macOS 14+ is required for MLX models according to LM Studio's system requirements.
Custom MLX Python runtimesSupported if they implement this architectureWorks for runtimes that can read MLX safetensors, Qwen3.5 MoE configs, the chat template, and MLX affine quantized language + vision tensors.
mlx-lm aloneText-side only / not sufficient for full VLM usemlx-lm is useful in the MLX ecosystem, but full image input needs the VLM path and processor stack.
Hugging Face Transformers / vLLM / TGINot directly loadableThese engines do not load this MLX quantized artifact directly. Use the original/full-precision model or produce a separate backend-specific quant.
Ollama / llama.cpp / KoboldCppNot directly loadableThese generally expect GGUF for local quantized inference. This repo is MLX safetensors, not GGUF.

Practical expectation: use this model on high-memory Apple Silicon Macs through MLX-VLM-compatible tooling. For image input, use PNG, JPEG, WebP, and PDF/image workflows supported by the serving app or preprocessing pipeline.

Research And Safety Notice

Why is this model Abliterated?

This model is intended for research, local experimentation, red-team evaluation, and authorized security testing. It may produce content that aligned models normally refuse. Users are responsible for applying appropriate safeguards and complying with laws and platform policies.

This is an abliterated model released for research and development, model-behavior analysis, authorized security testing, and experimentation with local Apple Silicon inference. Abliterated models may respond differently from aligned instruction models. Users are responsible for complying with applicable laws, platform policies, and safety requirements. The authors and uploaders are not responsible for misuse, harm, or unlawful deployment.

Reproducibility

The included mlztq_manifest.json records the source path, quantization recipe, weight format, vision quantization policy, and runtime contract used for this artifact. The benchmark files under benchmarks/results/ record the exact gate rows used to choose this model.