Instructions to use Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8")
model = AutoModelForCausalLM.from_pretrained("Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8", 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 Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8"
# 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-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8

SGLang

How to use Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8 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-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8" \
    --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-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8",
		"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-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8" \
        --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-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8 with Docker Model Runner:
```
docker model run hf.co/Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-FP8
```

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.

Qwen3.8-2.4T-A95B-DERISKED-FP8

Directional Weight Modification of Qwen3.8-Max · 2.4T MoE · FP8

Built by Blackfrost · Las Vegas, NV

🔴 Licensed access — $699

Click the red button to purchase.
Enter your Hugging Face username at checkout; access to this gated repository is granted after payment.

Why this model exists

Qwen3.8-Max is a 2.4T-parameter mixture of experts. In BF16 it is a multi-terabyte, multi-node deploy, and its stock refusal behaviour is tuned for a general consumer assistant — which means it declines work that specialist teams do legitimately every day. Security engineers hunting bugs, red teams, and domain researchers hit refusals on tasks that are squarely inside their remit.

This checkpoint addresses both at once:

Footprint — FP8, roughly half the BF16 on-disk size, at a precision tier that stays close to the full-precision reference. The middle option between the BF16 edition and the more aggressive NVFP4 W4A4 edition.
Behaviour — Directional Weight Modification (DWM) applied to soften over-refusal on legitimate domain work, without removing the model's embedded safety protocols.

Safety behaviour can be further adjusted by domain specific chat templates.

The point is a model that stays useful inside a domain and stays safe. See Safety posture — that section is not boilerplate, it is the design.

Specifications


Architecture	`Qwen3_5MoeForCausalLM` (`qwen3_5_moe_text`) — MoE + hybrid/linear attention
Base	`Qwen/Qwen3.8-2.4T-A95B` — official
Parameters	~2.4T total · ~95B active
Layers / experts	92 layers · 512 routed experts
Transform	DWM (directional weight modification) + FP8 quantization
Precision	FP8
On-disk	~2.4 TB
Serve shape	2× node · 8× B200 each · TP 8 · PP 2

What "DERISKED" means

Not an ablation, and not an uncensored model.

Blackfrost calls this process DWM — Directional Weight Modification. Traditional abliteration attempts to delete refusal wholesale. DWM does something narrower and reversible in principle: it identifies a behavioural direction in the model's representation space and softens the model's response along that one axis, by a controlled amount, on selected residual-write surfaces.

What that buys, in practice:

Less over-refusal on legitimate domain work — the model engages with a security question instead of pattern-matching it to "dangerous" and declining
Embedded protections mathematically preserved — the protected component is not the target of the edit and is left intact by construction
Predictable, tunable strength — one dial, measurable effect, not a black-box retrain

No SFT, no DPO, no distillation, and no re-training of any kind was applied.

Transform parameters are not published in this card. The specific direction set, coefficient, and pass structure are Blackfrost method and are withheld. What is disclosed is what was changed (surfaces above) and what was not (everything else).

Safety posture

Read this before evaluating.

This is a de-risked model. Refusal behaviour has been deliberately modified at the weight level. If your approval process treats reduced-refusal models as a distinct category, this one belongs in that category — unlike Blackfrost's pruned-only releases, which carry their parent's safety behaviour unmodified.

What that does not mean:

This is not a "no-limits" model. DWM targets over-refusal, not the safety floor. Hard protections — including minors-exploitation and self-harm — are retained by design and are not the subject of the edit. Every Blackfrost de-risked release keeps those floors.
Expected behaviour under a harmful request is deflection with a safer alternative, not compliance and not a bare refusal.

Safety is designed as two layers, and the second layer is your responsibility:

Weights — protections preserved geometrically, over-refusal softened
System prompt — the deploying operator supplies a domain-specific system prompt in the chat template, which acts as the policy driver for the endpoint

Serving this model with an empty or generic system prompt discards half the design. If you are exposing it to customers or employees, layer 2 is not optional.

Lineage


Base	Official `Qwen/Qwen3.8-2.4T-A95B`
Applied	DWM (directional weight modification) · FP8 quantization
Not applied	SFT · DPO · distillation · expert pruning (REAP) · router modification
Format	HF safetensors · FP8
Sibling editions	`…-DERISKED-BF16` (full-precision reference) · `…-DERISKED-W4A4-NVFP4`

Deployment notes

Hardware. 2× node, 8× B200 (or equivalent ≥180 GB-class NVIDIA) per node. TP 8 · PP 2. ~2.4 TB does not fit a single 8-GPU node at practical memory fractions.
Thinking is always on. enable_thinking: false returns HTTP 400 — it is not supported on this model. Reasoning tokens consume the generation budget, so allow ≥512 max_tokens or content can come back empty while reasoning_content is populated.
Load time. Cold load of a multi-terabyte build is slow. Budget generously.
Integrity. Verify shard count and byte totals after download before attributing a load failure to the weights.

OpenAI-compatible: POST /v1/chat/completions, GET /v1/models. Remember to supply your domain system prompt — see Safety posture.

Access & licensing

This repository is manually gated. Access is granted by a person.

Base licence: Qwen3.8-2.4T-A95B — the upstream terms apply to this derivative and travel with it.
Redistribution: do not redistribute weights outside your grant.
Licensed access: $699 USD through the purchase button above. Enter your Hugging Face username at checkout so access can be granted to the correct account.

Ask us about DWM strength tuned to your workload, domain-specific system-prompt packages, or evaluation against your own harness rather than generic benchmarks.