Instructions to use emwesoft/GLM-5.3-NVFP4-MTP-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 emwesoft/GLM-5.3-NVFP4-MTP-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 emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

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 emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./llama-cli -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

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 emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./build/bin/llama-cli -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

Use Docker

docker model run hf.co/emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

LM Studio
Jan

vLLM

How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "emwesoft/GLM-5.3-NVFP4-MTP-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": "emwesoft/GLM-5.3-NVFP4-MTP-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

Ollama
How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with Ollama:
```
ollama run hf.co/emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
```
Unsloth Desktop

How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

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": "emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4
```

Lemonade

How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

Run and chat with the model

lemonade run user.GLM-5.3-NVFP4-MTP-GGUF-NVFP4

List all available models

lemonade list

Hermes Agent

How to use emwesoft/GLM-5.3-NVFP4-MTP-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 emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

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 emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use emwesoft/GLM-5.3-NVFP4-MTP-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4

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 "emwesoft/GLM-5.3-NVFP4-MTP-GGUF:NVFP4" \
  --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"

GLM-5.3 753B — NVFP4 GGUF with MTP head

Native GGUF of incoai/GLM-5.3-NVFP4, the vendor's ModelOpt NVFP4 repack of zai-org/GLM-5.3, including block 78 (the MTP / NextN head) so the checkpoint's own head can drive speculative decoding. The NVFP4 trunk is kept as NVFP4 — not a requantisation.

465 GB, 11 shards. block_count 79, nextn_predict_layers 1, 1974 tensors, 225 NVFP4 tensors (identical to the no-MTP file — the trunk is not upcast).

Using the MTP head — read this

blk.78's experts are BF16, not NVFP4: 27 tensors, 18.54 GiB, roughly 3.7x a normal NVFP4 routed bank (5.06 GiB). If you use -ot with a CPU catch-all, pin block 78 to a GPU before the catch-all — -ot is first-match-wins, so otherwise blk.78 lands on CPU and every drafted token pays a CPU MoE pass, which is worse than not speculating.

-ot 'blk\.(...|78)\.ffn_.*_exps.*=CUDA1,\.ffn_.*_exps.*=CPU'
--spec-type draft-mtp --spec-draft-n-max 3

Related repos


Without MTP	`emwesoft/GLM-5.3-NVFP4-GGUF` — 20 GB smaller, no block 78
DFlash2 drafters	`emwesoft/GLM-5.3-DFlash2-GGUF` — the alternative to the MTP head

Engine requirements

llama.cpp with glm-dsa + GGML_TYPE_NVFP4. Three fixes are not yet upstream:

jinja numeric attribute access (obj.0) — GLM-5.3's chat template uses m.content.0.output. Without it the template throws, caps_get() swallows it, supports_tool_calls reports false, and every tool call comes back as plain text.
glm-dsa layer-input exposure — DFlash needs res->t_layer_inp[il]; without it attaching a drafter aborts on the first decode with GGML_ASSERT(t_layer_inp[il] != nullptr).
Whitespace tolerance before </tool_call> — a stray newline makes the streaming parser recognise a tool call then lose it, aborting from compute_diffs.

Measured throughput

2x RTX PRO 6000 Blackwell + 4x RTX 3090 + 251 GB RAM, 400K context, -t 36 -tb 40, experts partly CPU-resident (the weights do not fit in 288 GB of VRAM):

config	acceptance	decode
MTP head, `n-max 3`	74.9% (mean len 3.24)	9.3-12.2 tok/s
DFlash2 Q8_0, `n-max 4`	67.7% (mean len 3.69)	7.6-12.8 tok/s
no speculation	-	~10 tok/s

Throughput is prompt-dependent because acceptance is. Threads matter: on a 24-core/48-thread CPU, -t 48 collapsed decode to 0.5 tok/s — the ggml threadpool busy-spins and starves the CUDA submission thread. Leave headroom.

Sampling

From generation_config.json: temperature 1.0, top_p 0.95. The template exposes low/high/max reasoning effort only; anything else becomes max, and thinking cannot be disabled.

Downloads last month: -

GGUF

Model size

2B params

Architecture

dflash

Hardware compatibility

4-bit

8-bit

16-bit

Model tree for emwesoft/GLM-5.3-NVFP4-MTP-GGUF

Base model

zai-org/GLM-5.3

Quantized

incoai/GLM-5.3-NVFP4

Quantized

(2)

this model