Instructions to use TNSA/NGen-4OW-10T-Expiremental with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TNSA/NGen-4OW-10T-Expiremental with Transformers:

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

pipe = pipeline("text-generation", model="TNSA/NGen-4OW-10T-Expiremental", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("TNSA/NGen-4OW-10T-Expiremental", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use TNSA/NGen-4OW-10T-Expiremental with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TNSA/NGen-4OW-10T-Expiremental"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TNSA/NGen-4OW-10T-Expiremental",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/TNSA/NGen-4OW-10T-Expiremental

SGLang

How to use TNSA/NGen-4OW-10T-Expiremental 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 "TNSA/NGen-4OW-10T-Expiremental" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TNSA/NGen-4OW-10T-Expiremental",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "TNSA/NGen-4OW-10T-Expiremental" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TNSA/NGen-4OW-10T-Expiremental",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use TNSA/NGen-4OW-10T-Expiremental with Docker Model Runner:
```
docker model run hf.co/TNSA/NGen-4OW-10T-Expiremental
```

!!The Model Weights are still under Uploading stage!!

NGen-4OW-10T-Expiremental

Experimental research checkpoint package for the ngen4ow_10t architecture.

This repository is a derivative/repackaged research artifact initialized from moonshotai/Kimi-K2.6 weights. It reorganizes the source checkpoint into a top-level 10-layer routed layout:

ngen4ow_10t
  layers.0.expert_model
  layers.1.expert_model
  ...
  layers.9.expert_model

The intended runtime behavior is one active full-model layer per sequence. The logical package therefore advertises 10T total parameters and 32B active parameters. This is not an independently trained 10T model.

Direct vLLM serving requires custom ngen4ow_10t support.

Provenance

Source/base weights: moonshotai/Kimi-K2.6. Please also review the upstream model card and license terms before use.

Downloads last month: 744