Qwen3.8-27B

EXL3 4 bpw · sglang · 160K context · RTX 3090 Ti · 24 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.

130 tok/sdecode

776 tok/sprefill

160Kcontext window

Tested on this cardSep 28, 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:8000/v1.

1 Download the weights

hf download turboderp/Qwen3.8-27B-exl3 \
  --revision 4acd9ad5af224ec9e8815a54d71a033f378a7e9d \
  --local-dir ~/models/Qwen3.8-27B-exl3-4acd9ad5

2 Start the server

docker run --rm \
  --gpus all \
  -p 8000:30000 \
  --shm-size 16g \
  -e CUDA_DEVICE_ORDER=PCI_BUS_ID \
  -e HF_HUB_OFFLINE=1 \
  -e SGLANG_EXL3_EMBED_HOST=1 \
  -e SGLANG_EXL3_KERNEL=auto \
  -e SGLANG_EXL3_MODEL_PATH=/models/Qwen3.8-27B-exl3-4acd9ad5 \
  -v ~/models/Qwen3.8-27B-exl3-4acd9ad5:/models/Qwen3.8-27B-exl3-4acd9ad5:ro \
  --entrypoint /opt/entrypoint.sh \
  ghcr.io/0xsero/sglang-exl3:v0.7.0-multiarch@sha256:9d95829ab41a9307e9b25c86e4af1ec52c6a02f7888f0e13a1ce93c0a7ff3814 \
  python3 \
  -m sglang.launch_server \
  --model-path /models/Qwen3.8-27B-exl3-4acd9ad5 \
  --quantization exl3 \
  --trust-remote-code \
  --host 0.0.0.0 \
  --port 30000 \
  --served-model-name Qwen3.8-27B-exl3-4acd9ad5 \
  --context-length 163840 \
  --mem-fraction-static 0.92 \
  --kv-cache-dtype fp8_e4m3 \
  --max-running-requests 1 \
  --chunked-prefill-size 2048 \
  --max-prefill-tokens 2048 \
  --reasoning-parser qwen3 \
  --tool-call-parser qwen3_coder \
  --max-total-tokens 163840 \
  --mamba-ssm-dtype bfloat16 \
  --max-mamba-cache-size 5 \
  --cuda-graph-max-bs-decode 1 \
  --cuda-graph-bs-prefill 4 \
  8 \
  16 \
  32 \
  64 \
  128 \
  256 \
  512 \
  --speculative-algorithm NEXTN \
  --speculative-num-steps 3 \
  --speculative-eagle-topk 1 \
  --speculative-num-draft-tokens 4 \
  --speculative-token-map /opt/sglang-exl3/tokenmaps/qwen38_hot32k_v2.pt

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 RTX 3090 Ti on Sep 28, 2026.

Details

Weights: turboderp/Qwen3.8-27B-exl3 @ 4acd9ad5af ›
Image: ghcr.io/0xsero/sglang-exl3:v0.7.0-multiarch@sha256:9d95829ab41a9307e9b25c86e4af1ec52c6a02f7888f0e13a1ce93c0a7ff3814
Engine profile: sglang-exl3-multiarch ›
Recipe file: registry/recipes/nvidia/rtx-3090-ti-24gb/qwen3.8-27b.sglang.160k.json ›
Model released: Aug 5, 2026
Tested: vast, RTX 3090 Ti