Instructions to use tencent/DeepSeek-V3.1-Terminus-W4AFP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tencent/DeepSeek-V3.1-Terminus-W4AFP8 with Transformers:

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

pipe = pipeline("text-generation", model="tencent/DeepSeek-V3.1-Terminus-W4AFP8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tencent/DeepSeek-V3.1-Terminus-W4AFP8", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("tencent/DeepSeek-V3.1-Terminus-W4AFP8", trust_remote_code=True)
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use tencent/DeepSeek-V3.1-Terminus-W4AFP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tencent/DeepSeek-V3.1-Terminus-W4AFP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tencent/DeepSeek-V3.1-Terminus-W4AFP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tencent/DeepSeek-V3.1-Terminus-W4AFP8

SGLang

How to use tencent/DeepSeek-V3.1-Terminus-W4AFP8 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 "tencent/DeepSeek-V3.1-Terminus-W4AFP8" \
    --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": "tencent/DeepSeek-V3.1-Terminus-W4AFP8",
		"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 "tencent/DeepSeek-V3.1-Terminus-W4AFP8" \
        --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": "tencent/DeepSeek-V3.1-Terminus-W4AFP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tencent/DeepSeek-V3.1-Terminus-W4AFP8 with Docker Model Runner:
```
docker model run hf.co/tencent/DeepSeek-V3.1-Terminus-W4AFP8
```

DeepSeek-V3.1-Terminus-W4AFP8

This model is a mixed-precision quantized version of DeepSeek-V3.1-Terminus, with dense layer keep the FP8 quantization of the original model, while MoE layers uses INT4 weights and FP8 activation, also called W4AFP8.

Benchmark

The accuracy below was obtained with SGLang V0.5.3 in non-thinking mode.

Model	math_500	gpqa	aime2024	mmlu-pro
DeepSeek-V3.1-Terminus-W4AFP8	89.83	78.28	80.0	83.66

Inference with SGLang

We have already supported deploying this model using tensor parallel in sglang for better performance. The releated PR https://github.com/sgl-project/sglang/pull/8118 has been merged in SGLang V0.5.2, so you can deploy this model using SGLang version 0.5.2 or later with tensor parallel.

python3 -m sglang.launch_server --model-path /path/to/DeepSeek-V3.1-Terminus-W4AFP8 --tp 8 --trust-remote-code --host 0.0.0.0 --port 8000

Downloads last month: 525

Safetensors

Model size

349B params

Tensor type

F32

BF16

F8_E4M3

Model tree for tencent/DeepSeek-V3.1-Terminus-W4AFP8

Base model

deepseek-ai/DeepSeek-V3.1-Base

Quantized

deepseek-ai/DeepSeek-V3.1-Terminus

Finetuned

(5)

this model