Instructions to use Soofi-Project/Soofi-S-Base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Soofi-Project/Soofi-S-Base with Transformers:

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

pipe = pipeline("text-generation", model="Soofi-Project/Soofi-S-Base")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Soofi-Project/Soofi-S-Base", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Soofi-Project/Soofi-S-Base with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Soofi-Project/Soofi-S-Base

SGLang

How to use Soofi-Project/Soofi-S-Base 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 "Soofi-Project/Soofi-S-Base" \
    --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": "Soofi-Project/Soofi-S-Base",
		"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 "Soofi-Project/Soofi-S-Base" \
        --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": "Soofi-Project/Soofi-S-Base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Soofi-Project/Soofi-S-Base with Docker Model Runner:
```
docker model run hf.co/Soofi-Project/Soofi-S-Base
```

Obsolete before its released

by Scriptease - opened 9 days ago

Discussion

Scriptease

9 days ago

Hi,

it has been months which in ai terms feels like a year.

When will this be released or will it be obsolete before anyone can actually use it?

Kind regards

M-vdBerg

8 days ago

On 3 October, Aleph Alpha released Kolibri-1: a German/English MoE reasoning model, open weights, downloadable the day it was announced. German team, German-focused tokenizer, published limits.

Soofi-S-Base is still a gated closed-beta preview, the beta is closed to new applicants, and my question in #6 about a target month or quarter has had no answer in three weeks.

I’ll be honest: this is a disappointment. I wanted this project to be the one. The research direction is sound and the technical report is good work and that’s exactly what makes the silence hard to accept.
A homepage promising a release “this year” against a detail page saying “to be announced,” benchmarks published for a model nobody outside can run, and now two open threads asking the same question with no reply.
One sentence naming the remaining milestones would have kept me waiting.

I can’t evaluate what I can’t run.
I’m moving the sovereign-AI work I had planned for Soofi over to Kolibri-1...

kno10

6 days ago

It will be released when it's completely outdated, and they have finished the next one already to "beta test", and began yet another attempt.
Just as with Teuken / OpenGPT-X. They have this history of delaying releases as long as possible, if even delivering anything (c.f., EuroLingua-GPT with funding Timeline May 1, 2024 – October 1, 2025 that still has not released anything).

Use qwen3.8-27b or any successor. It is much better.
In contrast to European buzzword bingo, China delivers.

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