Qwen
/

Qwen3.8-2.4T-A95B

Text Generation

qwen3_5_moe_text

Model card Files Files and versions

Instructions to use Qwen/Qwen3.8-2.4T-A95B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Qwen/Qwen3.8-2.4T-A95B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Qwen/Qwen3.8-2.4T-A95B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3.8-2.4T-A95B")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.8-2.4T-A95B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Qwen/Qwen3.8-2.4T-A95B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Qwen/Qwen3.8-2.4T-A95B"
# 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/Qwen3.8-2.4T-A95B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Qwen/Qwen3.8-2.4T-A95B

How to use Qwen/Qwen3.8-2.4T-A95B 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/Qwen3.8-2.4T-A95B" \
    --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/Qwen3.8-2.4T-A95B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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/Qwen3.8-2.4T-A95B" \
        --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/Qwen3.8-2.4T-A95B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Qwen/Qwen3.8-2.4T-A95B with Docker Model Runner:
```
docker model run hf.co/Qwen/Qwen3.8-2.4T-A95B
```

Resources

View closed (4)

Can I use llama.cpp tools to quantize this model ONE shard at a time?

#43 opened 2 days ago by

Question

#42 opened 21 days ago by

smilyai-large-team

Architectural Proposal for Future Qwen Models: Dynamic Depth-Scaling and Macro-Expert Modularity for Composable AI

#41 opened about 1 month ago by

Qwen3.8 Max 0902?

#40 opened about 1 month ago by

Add Toolathlon-Verified evaluation result

#39 opened about 1 month ago by

Add Terminal-Bench evaluation results

#38 opened about 2 months ago by

ignore what’s going on in #26

#37 opened about 2 months ago by

Hellomaniamcoollol

notice how they all do good at around 2t

#36 opened about 2 months ago by

I researched the practical vLLM configs for Qwen3.8-2.4T-A95B — a few findings that surprised me more...

#35 opened about 2 months ago by

Qwen3.8 16B

#34 opened about 2 months ago by

Хочу gguf на 4_K_M квантования

#33 opened about 2 months ago by

Qwen3.8-A50B: Can We Distill A95B into a 50B MoE with 5–12B Active Parameters?

#32 opened about 2 months ago by

can you do 99999999999999B-A0.00000000000001B

#31 opened about 2 months ago by

Weirdos

#30 opened about 2 months ago by

I just checked Qwen3.8-2.4T-A95B for singular matrices and exploded condition numbers

#29 opened about 2 months ago by

对阿里太失望了

#28 opened about 2 months ago by

已收录到 trending.md

#27 opened about 2 months ago by

wtf is alibaba doing 😭✌️

#26 opened about 2 months ago by

Hellomaniamcoollol

Small models requests

#25 opened about 2 months ago by

Add WildClawBench evaluation result

#24 opened about 2 months ago by

建议延续马云的思想将千问团队剥离出来独立发展，借助资本做大做强

#23 opened about 2 months ago by

Looking forward to a ternary GGUF for this version

#22 opened about 2 months ago by

Small models are benchmark for the algorithm

#21 opened about 2 months ago by

Simply thank you :)

#19 opened about 2 months ago by

122b a10b?

#17 opened about 2 months ago by

We are waiting for 35B A3b model qwen 3.8! It will be the killer of all counterparts!

#15 opened about 2 months ago by

Not great, not horrible

#14 opened about 2 months ago by

Huge disappointment: Qwen 3.8 open weights are text-only and stripped of Qwen 3.8 Max features (No Vision, No 1M Context)

#13 opened about 2 months ago by

978 Downloads? 🤔

#12 opened about 2 months ago by

Why am I not suprised

#11 opened about 2 months ago by

Will this run on my phone?

#10 opened about 2 months ago by

Doctor-Chad-PhD

122b please

#9 opened about 2 months ago by

Text-only? Plans for multimodal release?

#7 opened about 2 months ago by

Add community evaluation results for DEEP-SWE, GPQA, HLE, SWE-BENCH_PRO

#6 opened about 2 months ago by

OMG

#5 opened about 2 months ago by

27B where

#4 opened about 2 months ago by

NEW MODEL

#3 opened about 2 months ago by

No 27B?

#2 opened about 2 months ago by

27b when?

#1 opened about 2 months ago by