Instructions to use unsloth/DeepSeek-V4-Flash-0731-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama cli -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama cli -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

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 unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./llama-cli -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

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 unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./build/bin/llama-cli -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Use Docker

docker model run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

LM Studio
Jan
Ollama
How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with Ollama:
```
ollama run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
```
Unsloth Desktop

How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with Docker Model Runner:
```
docker model run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL
```

Lemonade

How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Run and chat with the model

lemonade run user.DeepSeek-V4-Flash-0731-GGUF-UD-Q4_K_XL

List all available models

lemonade list

Hermes Agent

How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

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 unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use unsloth/DeepSeek-V4-Flash-0731-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q4_K_XL" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

pp sucks

#14

by gopi87 - opened Aug 1

Discussion

gopi87

Aug 1

i ahve b and ub at 6500 but i am getting 130 t/sec lol on pp and ts i am gett 7tps in latest llama.cpp in ik_llama.cpp its not even loading the ub and b what the fuck is going on this deep seek v4 flash arch shit

YearZero

Aug 1

skill issue

gopi87

Aug 1

skill issue

if you can why don't you help us

gopi87

Aug 1

echo 0 | sudo tee /proc/sys/kernel/numa_balancing
GGML_CUDA_NO_PINNED=1 CUDA_VISIBLE_DEVICES=2,3,0,1
numactl --interleave=all
~/llama.cpp/build/bin/llama-server
--model /mnt/nvme/deepseek/DeepSeek-V4-Flash-0731-UD-Q8_K_XL-00001-of-00005.gguf
--tensor-split 0.9,1.3,0.5,0.4
--reasoning on
--numa distribute
--no-mmap
--reasoning on
--n-cpu-moe 55
--ctx-size 150000
--batch-size 6500
--ubatch-size 6500
--parallel 1
--threads 45
--threads-batch 45
-ngl 100
--host 127.0.0.1
--port 8082
--jinja

this is what i am running

gopi87

Aug 2

echo 0 | sudo tee /proc/sys/kernel/numa_balancing
GGML_CUDA_NO_PINNED=1 CUDA_VISIBLE_DEVICES=2,3,0,1
numactl --interleave=all
~/llama.cpp/build/bin/llama-server
--model /mnt/nvme/deepseek/DeepSeek-V4-Flash-0731-UD-Q8_K_XL-00001-of-00005.gguf
--tensor-split 1.1,1.3,0,0.3
--override-tensor "blk.0.ffn_(up|down|gate)exps.weight=CUDA2,blk.1.ffn(up|down|gate)exps.weight=CUDA2,blk.2.ffn(down)_exps.weight=CUDA3"
--reasoning on
--numa distribute
--no-mmap
--n-cpu-moe 999
--flash-attn on
--kv-unified
--spec-type draft-dspark
--spec-draft-n-max 1
--ctx-size 280000
--batch-size 4000
--ubatch-size 4000
--verbose
--parallel 1
--threads 45
--threads-batch 45
-ngl 100
--host 127.0.0.1
--port 8082
--jinja

nothinking is improved this approches either

qenme

Aug 2

It's an issue of not being optimized for this new architecture. See here https://www.reddit.com/r/LocalLLaMA/comments/1vdm4z8/comment/p1aaigq/

gopi87

Aug 3

It's an issue of not being optimized for this new architecture. See here https://www.reddit.com/r/LocalLLaMA/comments/1vdm4z8/comment/p1aaigq/

i tested the build there build and solution also is not solved the pp issues its just stuck at 145 t/sec even though i sent it araound ub and b at 7300 values its just sucks

qenme

Aug 3

It's only faster if you have the full model in VRAM. In mainline I was getting 400-600 pp, with the fork I am getting 1300 pp.

It's an issue of not being optimized for this new architecture. See here https://www.reddit.com/r/LocalLLaMA/comments/1vdm4z8/comment/p1aaigq/

i tested the build there build and solution also is not solved the pp issues its just stuck at 145 t/sec even though i sent it araound ub and b at 7300 values its just sucks

It's only faster if you have the full model in VRAM. In mainline I was getting 400-600 pp, with the fork I am getting 1300 pp.

gopi87

Aug 3

•

edited Aug 3

It's only faster if you have the full model in VRAM. In mainline I was getting 400-600 pp, with the fork I am getting 1300 pp.

It's an issue of not being optimized for this new architecture. See here https://www.reddit.com/r/LocalLLaMA/comments/1vdm4z8/comment/p1aaigq/

i tested the build there build and solution also is not solved the pp issues its just stuck at 145 t/sec even though i sent it araound ub and b at 7300 values its just sucks

It's only faster if you have the full model in VRAM. In mainline I was getting 400-600 pp, with the fork I am getting 1300 pp.

ohh is the implementation for cpu+gpu moe types are still going btw with full vram 600 pf is not greate though

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