mlx-community/Qwen3.6-27B-OptiQ-4bit
Built with mlx-optiq, the MLX-native toolkit to quantize, fine-tune, and serve LLMs locally on Apple Silicon, no PyTorch and no cloud. Try the Lab · All OptiQ quants · Docs
A 4-bit mixed-precision MLX quant produced by mlx-optiq, the sensitivity-aware quantization toolkit for Apple Silicon. Beats stock uniform 4-bit on every benchmark in the six-metric Capability Score.
A 4-bit mixed-precision MLX quant of Qwen/Qwen3.6-27B. Per-layer bit-widths come from a KL-divergence sensitivity pass on a six-domain calibration mix (prose · reasoning · code · agent · tool-call · constraint-bearing instructions). Sensitive layers go to 8-bit; robust ones stay at 4-bit. The on-disk size is within ~5 % of a stock uniform 4-bit MLX quant.
Quantization details
| Property | Value |
|---|---|
| Predominant precision | 4-bit |
| Layers at 8-bit (sensitive) | 220 |
| Layers at 4-bit (robust) | 276 |
| Total quantized layers | 496 |
| Group size | 64 |
| Calibration mix | six-domain mix (40 samples × 6 domains) |
| Reference for sensitivity | bf16 (auto-resolved; falls back to uniform-4-bit if bf16 doesn't fit) |
| Bundled MTP head | mtp.safetensors (4-bit projections, BF16 norms), enables 1.4× decode via optiq serve --mtp |
We follow the same naming convention llama.cpp uses for Q4_K_M and similar mixed-precision quants: the "4-bit" label is for the predominant precision, not the weighted average. The mixed allocation is what lets this build beat stock uniform-4-bit on every benchmark below at the same disk size.
Usage
Load it with mlx-lm and use it as usual:
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/Qwen3.6-27B-OptiQ-4bit")
response = generate(
model, tokenizer,
prompt="Explain quantum computing in simple terms.",
max_tokens=200,
)
For more (mixed-precision KV-cache serving, sensitivity-aware LoRA fine-tuning, OpenAI + Anthropic-compatible inference server, hot-swap mounted adapters, sandboxed Python execution for agent workflows), install mlx-optiq:
pip install mlx-optiq
Speculative decoding (MTP)
This quant ships with a bundled Multi-Token Prediction head as mtp.safetensors. Enable it for ~1.4× faster decode:
optiq serve --model mlx-community/Qwen3.6-27B-OptiQ-4bit --mtp
Acceptance rate stays ~70% at depth 2 (the empirical sweet spot for Qwen3.6).
See the Qwen3.6 family guide on mlx-optiq.com for sampling defaults, training recipes, and family-specific caveats.
Benchmarks
Six-metric Capability Score (mean of MMLU + GSM8K + IFEval + BFCL + HumanEval + HashHop). Apples-to-apples comparison against stock uniform 4-bit:
| Metric | OptiQ | Uniform 4-bit | Δ |
|---|---|---|---|
| MMLU (5-shot, 1000 samples) | 87.4% | 87.6% | -0.2 |
| GSM8K (1000 samples, 3-shot CoT) | 92.0% | 92.1% | -0.1 |
| IFEval (full set, strict) | 74.1% | 71.7% | +2.4 |
| BFCL-V3 simple (200 calls) | 92.5% | 93.0% | -0.5 |
| HumanEval (164 problems, pass@1) | 90.2% | 92.1% | -1.8 |
| HashHop (long-context retrieval) | 80.0% | 77.0% | +3.0 |
| Capability Score (mean of 6) | 86.04 | 85.58 | +0.46 |
| KL vs uniform-4-bit reference (mean / p95) | 0.0895 / 0.4101 | , | , |
| On-disk size | 17.5 GB | 15.0 GB | +2.5 |
Every metric gets one equal vote. Disk size is reported next to the score as an honest second axis instead of being folded into the score. See the eval-framework writeup for the full methodology.
Links
- Project website: mlx-optiq.com
- Qwen3.6 family guide: mlx-optiq.com/docs/qwen3.6
- PyPI: pypi.org/project/mlx-optiq
- Calibration mix: mlx-optiq.com/blog/calibration-mix
- Eval framework: mlx-optiq.com/blog/eval-framework
- Base model: Qwen/Qwen3.6-27B
Quantize your own
This quant was produced by mlx-optiq. Point it at any Hugging Face model to get the same sensitivity-aware mixed precision:
pip install mlx-optiq
optiq convert <hf-model-id> --target-bpw 5.0 --candidate-bits 4,8
optiq lab # full local workbench: chat, compare, quantize, fine-tune
License
Apache 2.0 (inherits from base model).