Instructions to use antirez/deepseek-v4-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use antirez/deepseek-v4-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="antirez/deepseek-v4-gguf",
	filename="DeepSeek-V4-Flash-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use antirez/deepseek-v4-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf antirez/deepseek-v4-gguf:F32
# Run inference directly in the terminal:
llama-cli -hf antirez/deepseek-v4-gguf:F32

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf antirez/deepseek-v4-gguf:F32
# Run inference directly in the terminal:
llama-cli -hf antirez/deepseek-v4-gguf:F32

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf antirez/deepseek-v4-gguf:F32
# Run inference directly in the terminal:
./llama-cli -hf antirez/deepseek-v4-gguf:F32

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf antirez/deepseek-v4-gguf:F32
# Run inference directly in the terminal:
./build/bin/llama-cli -hf antirez/deepseek-v4-gguf:F32

Use Docker

docker model run hf.co/antirez/deepseek-v4-gguf:F32

LM Studio
Jan

vLLM

How to use antirez/deepseek-v4-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "antirez/deepseek-v4-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "antirez/deepseek-v4-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/antirez/deepseek-v4-gguf:F32

Ollama
How to use antirez/deepseek-v4-gguf with Ollama:
```
ollama run hf.co/antirez/deepseek-v4-gguf:F32
```

Unsloth Studio new

How to use antirez/deepseek-v4-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for antirez/deepseek-v4-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for antirez/deepseek-v4-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for antirez/deepseek-v4-gguf to start chatting

Pi new

How to use antirez/deepseek-v4-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf antirez/deepseek-v4-gguf:F32

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "antirez/deepseek-v4-gguf:F32"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use antirez/deepseek-v4-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf antirez/deepseek-v4-gguf:F32

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default antirez/deepseek-v4-gguf:F32

Run Hermes

hermes

Docker Model Runner
How to use antirez/deepseek-v4-gguf with Docker Model Runner:
```
docker model run hf.co/antirez/deepseek-v4-gguf:F32
```

Lemonade

How to use antirez/deepseek-v4-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull antirez/deepseek-v4-gguf:F32

Run and chat with the model

lemonade run user.deepseek-v4-gguf-F32

List all available models

lemonade list

Server doesnt seem to work

by sm54 - opened 9 days ago

Discussion

sm54

9 days ago

I've tried to load the q4 quant using ds4-server, and when I try to query it i dont get a response, nothing shows in the server window, so I'm thinking there's a connection issue? Although I can successfully retrieve the model list from it, it just wont return text when I send it a query.

antirez

Owner 9 days ago

Try ./ds4 command line utility to see if the inference is working. What Mac version / RAM amount? Thanks.

antirez

Owner 9 days ago

IMPORTANT: this quants ONLY WORK with DS4 inference engine

sm54

9 days ago

•

edited 9 days ago

It works if I run ds4 with a query, mac is an m3 ultra 256gb ram, running the q4 quant. Models are returned, but no text is.

Running the server using:

./ds4-server --ctx 81920 --kv-disk-dir /tmp/ds4-kv --kv-disk-space-mb 8192 --port 8072

sm54

9 days ago

Here's what I get when I run ds4

ds4 % ./ds4 -p "Explain Redis streams in one paragraph."
ds4: context buffers 1061.71 MiB (ctx=32768, backend=metal, prefill_chunk=2048, raw_kv_rows=2304, compressed_kv_rows=8194)
ds4: requesting Metal residency (may take tens of seconds)... done
ds4: warming Metal model views... done
ds4: Metal model views created in 3.255 ms, residency requested in 1017.287 ms, warmup 10.323 ms (mapped 157001.67 MiB from offset 5.08 MiB)
ds4: Metal mapped mmaped model as 1 overlapping shared buffers
ds4: Metal backend initialized for graph diagnostics
We need to explain Redis Streams in one paragraph. Keep it concise, covering key concepts: message queue, append-only log, consumer groups, persistence, etc.
Redis Streams is a data structure in Redis that functions as an append-only log, enabling reliable message streaming and event sourcing. It allows producers to add entries with unique IDs and field-value pairs, while consumer groups enable multiple consumers to collaboratively process messages with acknowledgment mechanisms, ensuring at-least-once delivery and fault tolerance. Streams support blocking reads, range queries, and trimming, making them ideal for building distributed message queues, real-time analytics, and event-driven architectures.
ds4: prefill: 30.91 t/s, generation: 31.69 t/s

If I run the server I get the below, but when I send a query nothing changes in the server window, and no response is received.

ds4 % ./ds4-server --ctx 81920 --kv-disk-dir /tmp/ds4-kv --kv-disk-space-mb 8192 --port 8072
ds4: requesting Metal residency (may take tens of seconds)... done
ds4: warming Metal model views... done
ds4: Metal model views created in 4.853 ms, residency requested in 1068.506 ms, warmup 6.194 ms (mapped 157001.67 MiB from offset 5.08 MiB)
ds4: Metal mapped mmaped model as 1 overlapping shared buffers
ds4: Metal backend initialized for graph diagnostics
0508 14:37:12 ds4-server: context buffers 2363.71 MiB (ctx=81920, backend=metal, prefill_chunk=2048, raw_kv_rows=2304, compressed_kv_rows=20482)
0508 14:37:12 ds4-server: KV disk cache /tmp/ds4-kv (budget=8192 MiB, cross-quant=accept, min=512, cold_max=30000, continued=10000, trim=32, align=2048)
0508 14:37:12 ds4-server: listening on http://127.0.0.1:8072

sm54

9 days ago

•

edited 9 days ago

I've asked GLM about it, it suggests that the lack of CORs support in the ds4-server code might be the issue? I get a response when I use curl but not when I send the request via open webui.

sm54

9 days ago

Have got the model to respond using your llama cpp fork instead of ds4-server, so am using that for now, thanks for your work.

sm54 changed discussion status to closed 9 days ago

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