Qwen3.5-9B

EXL3 4 bpw · tabbyapi · 256K context · RTX 5060 Ti · 16 GB

Alibaba's 9B model from February 2026: small enough for an 8 GB card, with a 64K–128K window. It thinks, calls tools and stays quick on modest hardware.

108 tok/sdecode

– tok/sprefill

256Kcontext window

Earlier checkSep 23, 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 TheMelonGod/Qwen3.5-9B-exl3 \
  --revision e98453aa143a24fab9eb14b718daed02e2fa6eef \
  --local-dir ~/models/Qwen3.5-9B-exl3-e98453aa

2 Write the server config

cat > qwen3.5-9b.tabbyapi.256k.yml <<'EOF'
network:
  host: 0.0.0.0
  port: 5000
  disable_auth: true
  disable_fetch_requests: true
  send_tracebacks: false
  api_servers: ["OAI"]
  sse_ping_interval: 15

logging:
  log_prompt: false
  log_generation_params: false
  log_requests: false
  log_chat_completion_requests: false

model:
  model_dir: /workspace/models
  inline_model_loading: false
  use_dummy_models: false
  model_name: Qwen3.5-9B-EXL3-6hb-4bpw
  backend: exllamav3
  max_seq_len: 262144
  cache_size: 262144
  cache_mode: Q4
  tensor_parallel: false
  gpu_split_auto: true
  autosplit_reserve: [192]
  chunk_size: 2048
  output_chunking: true
  max_batch_size: 2
  # Qwen3.5/3.8 think before answering; split that into reasoning_content instead of leaking it as text
  reasoning: true
  start_in_reasoning: auto

draft_model:
  draft_mode: mtp

sampling:
  override_preset:

memory:
  sysmem_recurrent_cache: 4096
  sysmem_kv_cache: 0
  cuda_malloc_async: true
EOF

3 Start the server

docker run --rm \
  --gpus all \
  -p 8000:5000 \
  --shm-size 8g \
  -e NVIDIA_VISIBLE_DEVICES=all \
  -v ~/models/Qwen3.5-9B-exl3-e98453aa:/workspace/models:ro \
  -v $PWD/qwen3.5-9b.tabbyapi.256k.yml:/app/config.yml:ro \
  --entrypoint /opt/venv/bin/python3 \
  ghcr.io/0xsero/tabbyapi-exl3@sha256:0f83e6198dc3be2561652d8df2525a7d1a69733e12fc8e25cbc5bc319a8a3ad1 \
  main.py \
  --config /app/config.yml

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 the older check (loads and chats); a full six-check run is pending.

Details

Weights: TheMelonGod/Qwen3.5-9B-exl3 @ e98453aa14 ›
Image: ghcr.io/0xsero/tabbyapi-exl3@sha256:0f83e6198dc3be2561652d8df2525a7d1a69733e12fc8e25cbc5bc319a8a3ad1
Launch: tabbyapi-qwen3.5-9b-4-bpw-6-bit-head-6hb-256k-v3 ›
Recipe file: registry/recipes/nvidia/rtx-5060-ti-16gb/qwen3.5-9b.tabbyapi.256k.json ›
Model released: Feb 26, 2026
Tested: earlier acceptance