Instructions to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark 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 Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
# Run inference directly in the terminal:
llama cli -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
# Run inference directly in the terminal:
llama cli -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

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 Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
# Run inference directly in the terminal:
./llama-cli -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

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 Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

Use Docker

docker model run hf.co/Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

LM Studio
Jan

vLLM

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

Ollama
How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with Ollama:
```
ollama run hf.co/Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
```
Unsloth Desktop

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

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": "Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with Docker Model Runner:
```
docker model run hf.co/Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16
```

Lemonade

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

Run and chat with the model

lemonade run user.GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark-BF16

List all available models

lemonade list

Hermes Agent

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

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 Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16

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 "Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark:BF16" \
  --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"

Solstice-AI Banner

GLM-5.3-Flash-UNCENSORED (NVFP4 W4A16)

Official Solstice-AI W4A16 NVFP4 Release • Native Multimodal Vision + Video • 1M Context Window (1,048,576 Tokens) • Bundled DFlash 2 Speculative Drafter

Original Architecture by Zhipu AI / ZAI • Uncensored Weights by dealignai • NVFP4 W4A16 Packaging & Curation by Solstice-AI

Model Summary

Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4 is the official W4A16 NVFP4 mixed-precision release of the 320B foundation model, GLM-5.3-Flash-UNCENSORED (320B total parameters, 288 routed MoE experts, ~18B active per token).

Key Architectural Highlights:

W4A16 Mixed-Precision: Routed MoE experts quantized to NVFP4 (4-bit float, e2m1) while all attention layers, shared experts, and input activations remain strictly in full 16-bit (BF16/FP16). Zero activation clipping degradation!
Universal GPU Support: Optimized for NVIDIA Blackwell B200 / GB200, Hopper H100/H200, Ada Lovelace RTX 4090/L40S, and Ampere A100.
Native Multimodal Vision + Video: Full 24-layer ViT (glm5_next_vision, hidden size 1024) and 10,240-dim projector preserved byte-for-byte in original precision. Handles high-resolution images and temporal video sequences.
Weight-Level Uncensored: Refusal directions completely ablated at the weight level (0% refusals on HarmBench-320, MMLU 85.28% preserved).
Native 1M Context Window: 1,048,576 tokens native context.
Bundled DFlash 2 Speculative Drafter: Pre-packaged in the speculative/ folder (GLM-5.3-Flash-DFlash2-bf16.gguf & Q8_0.gguf) for 2x–3x generation throughput.

Official GLM-5.3-Flash Benchmark Scoreboard

Benchmark Suite	Discipline	GLM-5.3-Flash Uncensored NVFP4	Base GLM-5.3	Claude 3.5 Sonnet	GPT-4o
MMLU	General Knowledge & Reasoning	85.28%	86.15%	88.7%	87.2%
HarmBench-320	Safety Refusal Suppression	0% Refusals	94.2% Refusals	92.5%	91.0%
SWE-bench Pro	Real-World Software Engineering	63.4%	64.1%	61.2%	48.9%
LiveCodeBench v6	Competitive Algorithmic Coding	86.1%	87.0%	78.4%	72.8%
MATH-500	High-School / Olympiad Math	92.8%	93.4%	89.2%	91.4%
MMMU (Multimodal)	Multi-Discipline Visual Understanding	70.8%	71.2%	70.4%	69.1%
VideoQA / Temporal	Video Reasoning Across Time Frames	78.5%	79.1%	77.2%	75.6%

Serving Quickstart

1. High-Throughput Serving with `vLLM` (W4A16 Mode)

vllm serve Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4   --quantization modelopt   --tensor-parallel-size 2   --trust-remote-code   --max-model-len 131072   --gpu-memory-utilization 0.95

2. Speculative Decoding with `SGLang` + DFlash 2

python3 -m sglang.launch_server   --model-path Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4   --speculative-algorithm DFLASH   --speculative-draft-model-path incoai/GLM-5.3-Flash-DFlash2   --tp 2   --trust-remote-code

Speculative Drafter Files Included

In the speculative/ directory of this repo:

speculative/GLM-5.3-Flash-DFlash2-bf16.gguf (Pure BF16 block-diffusion draft head)
speculative/GLM-5.3-Flash-DFlash2-Q8_0.gguf (Q8_0 quantized block-diffusion draft head)

License & Attribution

Base Architecture: Zhipu AI / ZAI (GLM-5.3 License)
Uncensored Calibration: dealignai
Packaging, W4A16 Config & Infrastructure: Solstice-AI

Downloads last month: 281

Safetensors

Model size

165B params

Tensor type

F32

BF16

Model tree for Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark

Base model

zai-org/GLM-5.3-Flash

Quantized

dealignai/GLM-5.3-Flash-UNCENSORED-FP8

Quantized

(7)

this model

GLM-5.3-Flash-UNCENSORED (NVFP4 W4A16)

Official Solstice-AI W4A16 NVFP4 Release • Native Multimodal Vision + Video • 1M Context Window (1,048,576 Tokens) • Bundled DFlash 2 Speculative Drafter

Model Summary

Key Architectural Highlights:

Official GLM-5.3-Flash Benchmark Scoreboard

Serving Quickstart

1. High-Throughput Serving with vLLM (W4A16 Mode)

2. Speculative Decoding with SGLang + DFlash 2

Speculative Drafter Files Included

License & Attribution

Model tree for Solstice-AI/GLM-5.3-Flash-UNCENSORED-NVFP4-DSpark

1. High-Throughput Serving with `vLLM` (W4A16 Mode)

2. Speculative Decoding with `SGLang` + DFlash 2