Instructions to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
# Run inference directly in the terminal:
llama cli -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
# Run inference directly in the terminal:
llama cli -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
# Run inference directly in the terminal:
./llama-cli -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

Use Docker

docker model run hf.co/puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

LM Studio
Jan

vLLM

How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "puwaer/DeepSeek-V4-Flash-0731-reap-200b-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": "puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

Ollama
How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with Ollama:
```
ollama run hf.co/puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
```
Unsloth Desktop

How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

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": "puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with Docker Model Runner:
```
docker model run hf.co/puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M
```

Lemonade

How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

Run and chat with the model

lemonade run user.DeepSeek-V4-Flash-0731-reap-200b-gguf-IQ1_M

List all available models

lemonade list

Hermes Agent

How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

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 puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M

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 "puwaer/DeepSeek-V4-Flash-0731-reap-200b-gguf:IQ1_M" \
  --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"

puwaer commited on Aug 26

Commit

c138be6

verified ·

1 Parent(s): 5397fac

Fix chat template: tool calling was broken and, more seriously, so was plain chat once tool support was wired up.

Browse files

What was broken:
- role: "tool" messages raised `unsupported role tool` in the template (reported as "MXFP4_MOE.gguf has wrong chat template").
- No `tools`/`tool_calls` support at all, so OpenAI-compatible clients that check template capabilities refused to forward `tools` (reported as "wrong template - no tool calls available").

What changed:
- Added tool role -> `<tool_result>` merging, `tools` schema rendering, and assistant `tool_calls` -> DSML rendering, ported 1:1 from the base model's `encoding/encoding_dsv4.py` (`merge_tool_messages`, `render_tools`, `encode_arguments_to_dsml`) and verified byte-exact against it over 300+ conversations including tool-call round trips.
- Along the way, fixed a template bug that would have broken *all* chat once tool support triggered llama.cpp's specialized DeepSeek V3.2/V4 response parser: the template never varied its output on `add_generation_prompt`, so llama.cpp's diff-based extraction of the generation prompt came back empty and every completion -- tool call or not -- was rejected as "does not match the expected peg-native format". Fixed by gating the trailing `<｜Assistant｜>` marker on `add_generation_prompt` only for the final message (mid-conversation transitions before a historical assistant turn stay unconditional, matching the base encoder, which has no such flag).

Verified end to end against a real llama-server (--jinja): plain chat in both thinking modes, a tool-result history round trip, and the model generating its own tool_calls, all succeed.

Files changed (1) hide show

DeepSeek-V4-Flash-0731-reap-200b-Q2_K.gguf +2 -2

DeepSeek-V4-Flash-0731-reap-200b-Q2_K.gguf CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad8f4021d4b1bd89a8d2f339bc2b0675bff9f4718263c220a321c4f1325259a4
-size 81409240736

 version https://git-lfs.github.com/spec/v1
+oid sha256:f8db89823dbaa141009e8c053b62c98a105d99f00b0f500b25f41ed5ddebe87c
+size 81409248512