Qwen3.8-27B

MLX 4-bit, native MTP · mtplx · 200K context · Apple M5 Pro · 48 GB

Alibaba's 27B dense model from August 2026, and the strongest model that fits one consumer card. It thinks before it answers, calls tools reliably and reads images, which makes it the default for coding agents and long projects.

22 tok/sdecode

179 tok/sprefill

200Kcontext window

Tested on this cardOct 6, 2026

Run it

Follow the steps below to download and run this recipe. Omarchy Local AI lists the recipes it can launch on your hardware. The server then answers on http://localhost:18198/v1.

1 Install the engine

python3 -m pip install annotated-doc==0.0.5 annotated-types==0.8.0 anyio==4.15.1 certifi==2026.7.22 click==8.5.0 fastapi==0.141.1 filelock==4.0.3 fsspec==2026.9.0 h11==0.16.0 hf-xet==1.6.0 httpcore==1.0.9 httpx==0.28.1 huggingface-hub==1.33.0 idna==3.20 jinja2==3.1.6 markdown-it-py==4.2.0 markupsafe==3.0.3 mdurl==0.1.2 mlx==0.32.2 mlx-lm==0.31.3 mlx-metal==0.32.2 mtplx==2.12.0 nanobind==3.1.0 numpy==2.5.3 packaging==26.3 pillow==12.3.0 protobuf==7.36.2 pydantic==2.13.5 pydantic-core==2.46.5 pygments==2.21.0 pyyaml==6.0.3 regex==2026.9.10 rich==15.0.0 safetensors==0.8.0 sentencepiece==0.2.2 shellingham==1.5.4 starlette==1.7.0 tokenizers==0.22.2 tqdm==4.70.1 transformers==5.14.1 typer==0.27.2 typing-extensions==4.16.0 typing-inspection==0.4.4 uvicorn==0.54.0

2 Download the weights

hf download Youssofal/Qwen3.8-27B-MTPLX-Bare-Speed \
  --revision b59d7002368575f08a58ba0d26686b53a9c162d6 \
  --local-dir models/Qwen3.8-27B-MTPLX-Bare-Speed-b59d7002

3 Start the server

HF_HUB_DISABLE_IMPLICIT_TOKEN=1 \
  HF_HUB_DISABLE_TELEMETRY=1 \
  HF_HUB_OFFLINE=1 \
  TOKENIZERS_PARALLELISM=false \
  TRANSFORMERS_OFFLINE=1 \
  mtplx \
  serve \
  --model models/Qwen3.8-27B-MTPLX-Bare-Speed-b59d7002 \
  --model-id qwen38-27b-mtplx-4bit \
  --host 127.0.0.1 \
  --port 18198 \
  --no-auth \
  --profile turbo \
  --generation-mode mtp \
  --depth 3 \
  --context-window 204800 \
  --kv-quant q4 \
  --prefill-chunk-tokens 4096 \
  --fan-mode default \
  --ssd-session-cache off \
  --agent-rewrites off \
  --no-stats-footer \
  --warmup-tokens 0 \
  --stream-stall-deadline-s 1800

What it passed

✓ loadLoads and serves within an hour

✓ chatAnswers a plain question and stops on its own

✓ reasoningThinks separately and gets 17 × 23 right

✓ toolsCalls a tool with the right arguments and uses the result

✓ contextRecalls a code buried in a prompt that fills 85% of the window

✓ speedDecodes at 15 tok/s or more

Passed all six checks on a real Apple M5 Pro 48GB (16-core GPU) on Oct 6, 2026.

Details

Weights: Youssofal/Qwen3.8-27B-MTPLX-Bare-Speed @ b59d700236 ›
Image: none: a program on the host
Launch: mtplx-qwen3.8-27b-4bit-mtp-200k ›
Recipe file: registry/recipes/apple/apple-m5-pro-48gb/qwen3.8-27b.mtplx.200k.json ›
Model released: Aug 5, 2026
Tested: owner, Apple M5 Pro 48GB (16-core GPU)