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"

New discussion

Resources

View closed (7)

DSpark enables DeepSeek to run 2x Faster!

pinned

🤗🔥 2

#36 opened about 1 month ago by

danielhanchen

KLD Benchmarks + why Q8_K_XL vs MXFP4 naming

pinned

❤️👍 11

#11 opened about 2 months ago by

danielhanchen

Run DeepSeek-V4-Flash-0731 in Unsloth! 🦥♥️

pinned

👍🔥 16

#6 opened about 2 months ago by

danielhanchen

My receipe 🍜 for Strix Halo 🤖

#49 opened 25 days ago by

auf1r2

Dynamic UD 3.0 GGUFs,,dear,please....

❤️ 2

#48 opened 26 days ago by

delphiR

The Model is pretty awsome but is it viable on my system? I appreciate any response!

#47 opened 27 days ago by

CHTL85

Any chance of a UD3.0 update?

🔥➕ 7

#46 opened about 1 month ago by

TizzyT566

Issues encountered when deploying UD-Q8_K_XL

#44 opened about 1 month ago by

dw1932

Can unsloth/DeepSeek-V4-Flash-0731-GGUF run on DwarfStar? Has anyone successfully run it? Could you please let me know?

#43 opened about 1 month ago by

mimeng1990

2x3090 Mixed GPU/CPU inference: -fit on and all good!

#42 opened about 1 month ago by

BingoBird

Q3_K_XL vs Q8_K_XL accuracy testing done anywhere?

#41 opened about 1 month ago by

jdchmiel

DAE know how to corretly run DSpark with part of the model offloaded to RAM using llama.cpp?

#40 opened about 1 month ago by

LukeC110

Are you sure the file for Q3_K_XL isn't mixed up with Q3_K_M?

#39 opened about 1 month ago by

inkor

Serving UD-IQ3_XXS on a 128 GB Strix Halo APU: measured config, and three settings that behave backwards here

#38 opened about 1 month ago by

wowo101

memory size context

#35 opened about 1 month ago by

nemilos

DeepSeek-V4-Flash-0731 on 6× Tesla V100 32GB – Q8_K_XL, DSpark and legacy MTP results

👍 3

#34 opened about 1 month ago by

sandrox5

DSpark: Inconsistency in num_speculative_tokens

#32 opened about 2 months ago by

Pragmatism0220

decoding with DSpark is slower on llama

#31 opened about 2 months ago by

SlavikF

Release imatrix

#30 opened about 2 months ago by

KeinNiemand

dspark

#29 opened about 2 months ago by

ludashi666

New GGUFs required? "chat : add new template for DeepSeek V4 Flash 0731 (#26398)"

👍 4

#28 opened about 2 months ago by

rtzurtz

How did you quantize without imatrix?

#27 opened about 2 months ago by

Wladastic

dspark-DeepSeek-V4-Flash-0731-Q8_0.gguf FAILS to load in the latest llama-server

➕ 1

#26 opened about 2 months ago by

Nerdsking

UD-IQ4_XS 137 GB or UD-IQ4_NL 137 GB?

👀 2

#25 opened about 2 months ago by

BingoBird

DeepSeek V4 @ IQ3XXS on M1 Ultra 128GB- 6 tok/s -> 16 tok/s in LM Studio after patch

#24 opened about 2 months ago by

jeevacation2

CUDA ERROR for omp

#23 opened about 2 months ago by

Butterfly-314

For DGX Spark (GB10) owners: a vLLM alternative to GGUF with the native DSpark speculative head - ~10 tok/s, 256K ctx, tool calling

#22 opened about 2 months ago by

wiklif

Why do the K quants use I quants?

#21 opened about 2 months ago by

LagOps

Success to run unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-Q8_K_XL with 2xA100/80G

#20 opened about 2 months ago by

weisunding

Great agentic results even with IQ1_S!

👍🤗 8

#19 opened about 2 months ago by

phakio

Best way to run on Dual RTX 6000 Pro Blackwell?

#18 opened about 2 months ago by

jhgoodwin

Running IQ2_M on an HP ZBook Ultra G1a (Strix Halo, 128 GB Unified Memory) at 11.49 tok/s with a 1M Context

🔥👍 4

#17 opened about 2 months ago by

AI-Joe-git

In mac it falls back to CPU. Any ideas?

👀 1

#16 opened about 2 months ago by

Grossor

pp sucks

#14 opened about 2 months ago by

gopi87

There’s an inexplicable VRAM hogging issue with this version of DeepSeek Flash compared to the previous one.

👍 2

#13 opened about 2 months ago by

Kosh69

3bit working, 4bit not

👍 4

#12 opened about 2 months ago by

tasticleeze

Dual RTX3090 and 64GB DDR4 + 78GB Swap. Speed: ~1.78 TPS

🤝👍 4

#8 opened about 2 months ago by

robert1968

Best inference engine to use at the moment for the full quant on 96gb vram

#4 opened about 2 months ago by

void009

Pulling UD_Q8 right now, MTP questions lol

#3 opened about 2 months ago by

Chuck572

Lossless UD-Q8_K_XL. UD-Q4_K_XL also provided for now

❤️🔥 15

#2 opened about 2 months ago by

danielhanchen

Awww yiss

🚀🔥 26

#1 opened about 2 months ago by

CurbStomper