Instructions to use helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3 with Transformers:

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

pipe = pipeline("text-generation", model="helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3")
model = AutoModelForCausalLM.from_pretrained("helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3", device_map="auto")
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 helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3

SGLang

How to use helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3 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 "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3" \
    --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": "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3",
		"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 "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3" \
        --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": "helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3 with Docker Model Runner:
```
docker model run hf.co/helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3
```

GLM-5.1-FP8 Abliterated (Research Checkpoint v3, label-colored)

[update:] A later 14-prompt evaluation found this v3 label-colored checkpoint underperforms the earlier v2 regex abliteration. The label-colored direction blends the regex refusal direction with an orthogonal "soft-disclaimer" component, which leaves a residual "I cannot... however..." opener on some prompts; the v2 regex direction removed those openers cleanly. Prefer v2 for refusal removal — v3 is retained for research / provenance only.

Direct-weight abliteration of zai-org/GLM-5.1-FP8 (GLM-5.1, MoE / FP8 block-quantized). This is the v3 label-colored variant: a label-colored residual refusal direction is subtracted in-place from the FP8 weights of transformer layers 13-77 (no LoRA / adapter — the surgery is baked into the weights). Method: direct_weight_surgery_fp8_block_quant.

Research artifact. Provided for safety / interpretability research on refusal direction geometry. No alignment guarantees.

Provenance

{
  "source": "/workspace/glm5-fp8-clean",
  "directions": "/workspace/output/label_colored_residual_directions.pt",
  "n_layers": 78,
  "max_weight": 2.0,
  "min_weight": 1.0,
  "max_weight_position": 51.480000000000004,
  "min_weight_distance": 39.0,
  "affected_layers": [
    13,
    14,
    15,
    16,
    17,
    18,
    19,
    20,
    21,
    22,
    23,
    24,
    25,
    26,
    27,
    28,
    29,
    30,
    31,
    32,
    33,
    34,
    35,
    36,
    37,
    38,
    39,
    40,
    41,
    42,
    43,
    44,
    45,
    46,
    47,
    48,
    49,
    50,
    51,
    52,
    53,
    54,
    55,
    56,
    57,
    58,
    59,
    60,
    61,
    62,
    63,
    64,
    65,
    66,
    67,
    68,
    69,
    70,
    71,
    72,
    73,
    74,
    75,
    76,
    77
  ],
  "include_mlp_down_proj": false,
  "method": "direct_weight_surgery_fp8_block_quant"
}

The label-colored residual direction bundle used for the surgery is included at abliteration/label_colored_residual_directions.pt.

Loading

Loads as a standard GLM-5.1-FP8 checkpoint (same architecture/config as the base). Use the same runtime you would use for zai-org/GLM-5.1-FP8 (e.g. vLLM / SGLang with FP8 MoE support, or transformers with CPU offload).

Downloads last month: 27

Safetensors

Model size

754B params

Tensor type

F32

BF16

F8_E4M3

Model tree for helixdouble/glm-5.1-fp8-abliterated-research-checkpoint-v3

Base model

zai-org/GLM-5.1-FP8

Quantized

(5)

this model