Instructions to use deepseek-ai/DeepSeek-V4.1-Flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepseek-ai/DeepSeek-V4.1-Flash with Transformers:

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

pipe = pipeline("image-text-to-text", model="deepseek-ai/DeepSeek-V4.1-Flash")

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-V4.1-Flash", device_map="auto")

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use deepseek-ai/DeepSeek-V4.1-Flash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deepseek-ai/DeepSeek-V4.1-Flash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deepseek-ai/DeepSeek-V4.1-Flash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/deepseek-ai/DeepSeek-V4.1-Flash

SGLang

How to use deepseek-ai/DeepSeek-V4.1-Flash 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 "deepseek-ai/DeepSeek-V4.1-Flash" \
    --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": "deepseek-ai/DeepSeek-V4.1-Flash",
		"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 "deepseek-ai/DeepSeek-V4.1-Flash" \
        --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": "deepseek-ai/DeepSeek-V4.1-Flash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use deepseek-ai/DeepSeek-V4.1-Flash with Docker Model Runner:
```
docker model run hf.co/deepseek-ai/DeepSeek-V4.1-Flash
```

DeepSeek-V4.1-Flash / inference /README.md

GeeeekExplorer

Update inference/README.md

2bc89ac verified 2 days ago

preview code

Raw

History Blame Contribute Delete

2.03 kB

Minimal inference

A readable reference implementation rather than a production serving engine. The model code covers the vision encoder and aligner, sliding-window plus compressed sparse attention with its two-level indexer, engram n-gram lookups, MoE, Hyper-Connections, and the DSpark forward path. Generation itself is plain autoregressive sampling.

Install

python -m pip install -r requirements.txt

Convert Hugging Face weights

The runtime uses one converted checkpoint file per tensor-parallel rank. From this directory:

export HF_CKPT_PATH=/path/to/DeepSeek-V4.1-Flash-HF
export SAVE_PATH=/path/to/DeepSeek-V4.1-Flash-TP8
export MP=8

python convert.py \
  --hf-ckpt-path "${HF_CKPT_PATH}" \
  --save-path "${SAVE_PATH}" \
  --model-parallel "${MP}" \
  --expert-dtype fp4 \
  --tokenizer-path "${HF_CKPT_PATH}"

Expert counts are inferred from the weight names, so they do not need to be passed. --tokenizer-path points at whichever directory holds tokenizer.json and tokenizer_config.json; they are copied into the converted checkpoint.

Run the equivalent TXT and JSON examples

export CKPT_PATH=/path/to/DeepSeek-V4.1-Flash-TP8
export MP=8

INPUT_FILE=examples/example.txt ./run.sh
INPUT_FILE=examples/example_harmony.json ./run.sh

The two files express the same interleaved two-image prompt, so they produce identical encoded prompts and input token IDs.

For interactive chat:

torchrun --nproc-per-node "${MP}" generate.py \
  --ckpt-path "${CKPT_PATH}" \
  --config config.json \
  --interactive \
  --temperature 0.6

For multi-node execution, pass the usual torchrun --nnodes, --node-rank, --master-addr, and --master-port arguments before generate.py.

Self-test

model.py builds a small model from the ModelArgs defaults and runs a prefill plus 22 decode steps, exercising the real dense-fp8 / MoE-fp4 kernels. Weights are uninitialized, so it checks shapes and kernel plumbing, not numerics:

python model.py