Qwen3.8-27B

EXL3 4 bpw · vllm · 256K context · Arc Pro B70 32 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.

82 tok/sdecode

775 tok/sprefill

256Kcontext window

Tested on this cardSep 25, 2026

Run it

With Omarchy Local AI it is one button. By hand, it is three steps: the weights, the config, the container. The server then answers on http://localhost:8000/v1.

1 Download the weights

hf download turboderp/Qwen3.8-27B-exl3 \
  --revision 113cf7ab958054860e43fb7f3063b1af19171095 \
  --local-dir ~/models/Qwen3.8-27B-exl3-113cf7ab

2 Start the server

docker run --rm \
  --device /dev/dri \
  -p 8000:8000 \
  --shm-size 32g \
  -e HF_HUB_OFFLINE=1 \
  -v ~/models/Qwen3.8-27B-exl3-113cf7ab:/models:ro \
  ghcr.io/0xsero/exl3xpu@sha256:5d8630ed6a1a066c5e51d853372df4a6c210ede7c4448d703c8bb669f7dcec3d \
  models/qwen3.8-27b-exl3-4.00bpw \
  --gpu 0 \
  --port 8000 \
  --model-path /models \
  -- \
  --enable-prefix-caching \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_coder \
  --reasoning-parser qwen3

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 Intel Arc Pro B70 (owner: omarchy box) on Sep 25, 2026.

Details

Weights: turboderp/Qwen3.8-27B-exl3 @ 113cf7ab95 ›
Image: ghcr.io/0xsero/exl3xpu@sha256:5d8630ed6a1a066c5e51d853372df4a6c210ede7c4448d703c8bb669f7dcec3d
Engine profile: vllm-qwen3.8-27b-exl3-4bpw-256k ›
Recipe file: registry/recipes/intel/intel-arc-pro-b70-32gb/qwen3.8-27b.vllm.256k.json ›
Model released: Aug 5, 2026
Tested: owner, Intel Arc Pro B70 (owner: omarchy box)