Instructions to use Qwen/Qwen3.5-35B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Qwen/Qwen3.5-35B-A3B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="Qwen/Qwen3.5-35B-A3B") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("Qwen/Qwen3.5-35B-A3B") model = AutoModelForMultimodalLM.from_pretrained("Qwen/Qwen3.5-35B-A3B") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Inference
- HuggingChat
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Qwen/Qwen3.5-35B-A3B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Qwen/Qwen3.5-35B-A3B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Qwen/Qwen3.5-35B-A3B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/Qwen/Qwen3.5-35B-A3B
- SGLang
How to use Qwen/Qwen3.5-35B-A3B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Qwen/Qwen3.5-35B-A3B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Qwen/Qwen3.5-35B-A3B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Qwen/Qwen3.5-35B-A3B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Qwen/Qwen3.5-35B-A3B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use Qwen/Qwen3.5-35B-A3B with Docker Model Runner:
docker model run hf.co/Qwen/Qwen3.5-35B-A3B
merge -> Agents-A1 / Ornith-1.0-35B / Qwen3.5-35B-A3B
π 1
#82 opened about 1 hour ago
by
tepirale
Listed on OpenModelMap
#81 opened 3 days ago
by
duola15
Listed on OpenModelMap
#80 opened 3 days ago
by
duola15
Listed on OpenModelMap
#79 opened 14 days ago
by
duola15
Add EvalEval community eval results
#78 opened 29 days ago
by
EvalEvalBot
Given a search question, return reasoning and content("\n\n") with no tool calls in first turn ending with finish_reason==stop.
#77 opened about 1 month ago
by
xingmo
Add MMMU-Pro evaluation result
#76 opened about 2 months ago
by
SaylorTwift
When will there be an Instruct version?
1
#75 opened 2 months ago
by
aklgh
[Docs] Conflicting sampling parameters in README.md
#74 opened 3 months ago
by
trgtszr
Add vLLM offline inference example for Qwen3.5 -- Added in README.md
#73 opened 3 months ago
by
hrithiksagar-bgen
Compact tool-rendering template β 49% fewer prompt tokens
#72 opened 3 months ago
by
Evita
[Bug] Jinja template preventing multiple system messages
#71 opened 3 months ago
by
phly95
Reddit Claim - training bug in Qwen3.5 35B A3B
1
#70 opened 3 months ago
by
jpsequeira
Some tokens in Qwen3.5 vocabulary cannot be generated from the merges file
#69 opened 3 months ago
by
jslhcl
fix chat template to avoid empty historical `<think>` blocks
4
#68 opened 3 months ago
by
latent-variable
Why does the thinking mode in Qwen3.5 35B MoE only output a closing </think> tag without an opening <think> tag?
π 1
3
#67 opened 3 months ago
by
Ttooong
Qwen3.5 - 35b some time skip the thinking stage then I don't disable thinking
#66 opened 3 months ago
by
ming514810
Non-deterministic outputs with identical inputs (GPTQ-Int4, vLLM 0.18, document extraction)
#65 opened 3 months ago
by
Starlight1997
Question about tool definition format in chat template: should tools be passed as JSON or XML?
π 1
1
#64 opened 3 months ago
by
SanJiaoMao
First MTP GGUF + custom quantization pipeline for this model β looking for arXiv endorser
#63 opened 3 months ago
by
Kevletesteur
llama.cpp load issue with this model.. GGUF
#62 opened 3 months ago
by
LinuxMagic
Fixed jinja template that was causing stalling and stopping the responses when using tools.
π 2
#61 opened 3 months ago
by
deladuck
Proposal: new chat_template_arg `enable_history_reasoning` for reusing prompt cache among querys within Agents .
#60 opened 3 months ago
by
Abioy
Add ScreenSpot-Pro evaluation result
#59 opened 4 months ago
by
merve
Add MathArena evaluation result for hmmt/hmmt_feb_2026
#58 opened 4 months ago
by
JasperDekoninck
Add MathArena evaluation result for aime/aime_2026
#57 opened 4 months ago
by
JasperDekoninck
21-fix chat template β addresses tool calling crash (discussion #4) and other open bugs
π 1
3
#56 opened 4 months ago
by deleted
Hugging Face vision model using question
#55 opened 4 months ago
by
Day1Kim
ζΊη₯οΌ
2
#54 opened 4 months ago
by
DGHMFFD
Question
1
#53 opened 4 months ago
by
y198
I paid $3,800 to build me a 30-agent AI swarm. It then moved into my house and stopped costing me money.
ππ 4
3
#52 opened 4 months ago
by
digiegold
Update README.md
1
#51 opened 4 months ago
by
cryptosxion
uh
#50 opened 4 months ago
by
cryptosxion
llama.cpp prompt reanalyze issue
2
#49 opened 4 months ago
by
mayankiit04
Forget this nonsnse
#48 opened 4 months ago
by
ZiggyS
good
1
#45 opened 4 months ago
by
pallabkr77
RIP
π€ 10
2
#44 opened 4 months ago
by
kabachuha
Benchmarks, tradeoffs, and live video inference across five Qwen 3.5 models
π₯ 6
#43 opened 4 months ago
by
younes-ovs
Learning to be able
#42 opened 4 months ago
by
Fairyydr0p
Model feedback: Problem with handling latest new articles
3
#41 opened 4 months ago
by
babytifa
Upload ChromeSetup.exe
#40 opened 4 months ago
by
cabot1
vLLM - Looping prevention
π 1
1
#39 opened 4 months ago
by
janreges3
One flaw in the architecture - has to reprocess the prompt every time
βπ 2
#38 opened 4 months ago
by
Dampfinchen
hello everyone
#37 opened 4 months ago
by
hoangminhkhoi
tempek
π 1
1
#36 opened 4 months ago
by
Funkist
Upload test.txt
#32 opened 4 months ago
by
Moodymue