Instructions to use swiss-ai/Apertus-70B-Instruct-2509 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use swiss-ai/Apertus-70B-Instruct-2509 with Transformers:

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

pipe = pipeline("text-generation", model="swiss-ai/Apertus-70B-Instruct-2509")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("swiss-ai/Apertus-70B-Instruct-2509")
model = AutoModelForCausalLM.from_pretrained("swiss-ai/Apertus-70B-Instruct-2509")
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use swiss-ai/Apertus-70B-Instruct-2509 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "swiss-ai/Apertus-70B-Instruct-2509"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "swiss-ai/Apertus-70B-Instruct-2509",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/swiss-ai/Apertus-70B-Instruct-2509

SGLang

How to use swiss-ai/Apertus-70B-Instruct-2509 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 "swiss-ai/Apertus-70B-Instruct-2509" \
    --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": "swiss-ai/Apertus-70B-Instruct-2509",
		"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 "swiss-ai/Apertus-70B-Instruct-2509" \
        --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": "swiss-ai/Apertus-70B-Instruct-2509",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use swiss-ai/Apertus-70B-Instruct-2509 with Docker Model Runner:
```
docker model run hf.co/swiss-ai/Apertus-70B-Instruct-2509
```

GPU-Setup

#14

by hideosnes - opened Dec 29, 2025

Discussion

hideosnes

Dec 29, 2025

What GPU setup do you recommend to run Apertus ideally on and what setup might still do the trick?

loleg

Swiss AI Initiative org Apr 21

Apertus is part of an R&D initiative, and the parameters of the technical deployment of our models is largely a matter of third party software and hardware choices. There is no general guidance on system requirements. Speaking from my own experience, it is possible to get the 8B model to run quite well on consumer-level hardware. 12 GB of VRAM is advisable, but quantized versions exist to fit much smaller amounts of memory. To get the full 70B version of Apertus to run, you will ideally need an industrial-strength graphics card. Again, it depends a lot on what kind of performance you expect to have, how many users you will serve etc. There are some reviews online, which may provide relevant performance (tokens / second) benchmarks. I would suggest contacting some of the providers that deploy Apertus, do your own experiments, and please share any useful advice that you hear about.

loleg changed discussion status to closed Apr 21

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment