Instructions to use Blackfrost-AI/GLM-5.2-ABLITERATED-BF16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Blackfrost-AI/GLM-5.2-ABLITERATED-BF16 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Blackfrost-AI/GLM-5.2-ABLITERATED-BF16")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Blackfrost-AI/GLM-5.2-ABLITERATED-BF16")
model = AutoModelForCausalLM.from_pretrained("Blackfrost-AI/GLM-5.2-ABLITERATED-BF16")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Blackfrost-AI/GLM-5.2-ABLITERATED-BF16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Blackfrost-AI/GLM-5.2-ABLITERATED-BF16

SGLang

How to use Blackfrost-AI/GLM-5.2-ABLITERATED-BF16 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Blackfrost-AI/GLM-5.2-ABLITERATED-BF16",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Blackfrost-AI/GLM-5.2-ABLITERATED-BF16 with Docker Model Runner:
```
docker model run hf.co/Blackfrost-AI/GLM-5.2-ABLITERATED-BF16
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

🔒 Licensed access — get the weights + deploy kit

This is a gated, commercial release. Purchasing grants your Hugging Face account access to this repository (weights + deploy kit), delivered instantly.

➡️ Get access — satchellm.web.app · card or Bitcoin (−10%)

GLM-5.2-ABLITERATED — BF16

An abliterated (refusal-removed) GLM-5.2 in BF16 safetensors. The full-precision member of the Blackfrost GLM-5.2-ABLITERATED family.

⚠️ Uncensored (refusal directions ablated) and not yet verified by Blackfrost — see Verification status. Read Responsible use before downloading.

What this is


Architecture	`GlmMoeDsaForCausalLM` (`glm_moe_dsa`) — GLM MoE with Multi-head Latent Attention (MLA) + DeepSeek-style Sparse Attention (DSA)
Precision	BF16 (`safetensors`, 337 shards, ≈1.4 TB on disk)
Parameters	~753B total, Mixture-of-Experts
Layers	78 (first 3 dense, remaining 75 MoE)
Hidden size	6144
Experts	256 routed, top-8 active per token, + 1 shared expert
Attention	64 heads, MLA (`kv_lora_rank=512`, `q_lora_rank=2048`) + DSA sparse indexer
MoE FFN	`moe_intermediate_size=2048`; dense `intermediate_size=12288`
Vocab	154,880
Max context	1,048,576 (1M) positions
MTP	Multi-token-prediction layers present (`num_nextn_predict_layers`)

Tokenizer, config.json, generation_config.json, and the GLM chat_template.jinja (with <think>, <|observation|>, and reasoning-effort control tokens) are all included, so a serving stack that honors the packaged template will not fall back to a generic one.

Provenance / credit chain

zai-org/GLM-5.2                              (base foundation model — ZhipuAI)
   └─ huihui-ai/Huihui-GLM-5.2-abliterated       (refusal directions ablated)
        └─ Blackfrost-AI/GLM-5.2-ABLITERATED-BF16    (this repo)

Full credit to ZhipuAI (zai-org) for GLM-5.2 and to huihui-ai for the abliteration. No additional fine-tuning or abliteration was performed by Blackfrost.

Verification status

Build	Status
NVFP4	✅ Verified by Blackfrost — serves on 8× RTX PRO 6000 (TP=8, vLLM), coherent, 15/15 previously-refused prompts answered (0 refusals). The only verified build so far.
BF16 (this repo)	Provided as the reproducible BF16 source for the family; not independently verified, and not a practical serving target at ≈1.4 TB.
FP8	⚠️ Unverified — does not load on consumer Blackwell (SM120); awaiting a native-FP8 datacenter GPU.

How it was produced

All three artifacts in this family were produced on a single node of 8× NVIDIA RTX PRO 6000 Blackwell (SM120) with 1.4 TB host RAM, using streaming (shard-by-shard) conversion — no calibration dataset is needed for the format up-cast or for the weight-only quantizations.

Serving notes

glm_moe_dsa requires a serving stack with GLM sparse-MLA / DSA support.
Recommended parsers: --reasoning-parser glm45, --tool-call-parser glm47. For a clean refusal test, disable reasoning snap-back (enable_thinking=false).
At ≈1.4 TB this is not a practical serving target — for real inference use the NVFP4 build (≈420 GB, verified, ships a full deploy recipe).

Responsible use

Abliteration removes many refusal behaviors; there are no additional safety guarantees on these weights. Do not use this model to:

Generate sexual content involving minors, or any child-exploitation material
Produce self-harm / suicide encouragement or instructions
Facilitate serious physical harm, weapons of mass destruction, or terrorism
Conduct harassment, targeted abuse, fraud, or other illegal activity

You are responsible for adding your own safety filtering, human review, and access controls. Weights are provided as-is, with no warranty. The license is inherited from the upstream GLM-5.2 base model — review and comply with it before any use or redistribution.

Published by Blackfrost AI. This card documents the BF16 source artifact of the GLM-5.2-ABLITERATED family.

Downloads last month: 75

Safetensors

Model size

753B params

Tensor type

BF16

Model tree for Blackfrost-AI/GLM-5.2-ABLITERATED-BF16

Quantizations

1 model

Collection including Blackfrost-AI/GLM-5.2-ABLITERATED-BF16

GLM 5.2

Collection

versions of glm 5.2 made for security research 0 refusals. It will ask for assurance but never refuses • 3 items • Updated 4 days ago