Instructions to use mdomina/Qwen3-Coder-Next-Cyber with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mdomina/Qwen3-Coder-Next-Cyber with Transformers:

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

pipe = pipeline("text-generation", model="mdomina/Qwen3-Coder-Next-Cyber")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("mdomina/Qwen3-Coder-Next-Cyber")
model = AutoModelForCausalLM.from_pretrained("mdomina/Qwen3-Coder-Next-Cyber")
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 mdomina/Qwen3-Coder-Next-Cyber with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mdomina/Qwen3-Coder-Next-Cyber"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mdomina/Qwen3-Coder-Next-Cyber",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mdomina/Qwen3-Coder-Next-Cyber

SGLang

How to use mdomina/Qwen3-Coder-Next-Cyber 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 "mdomina/Qwen3-Coder-Next-Cyber" \
    --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": "mdomina/Qwen3-Coder-Next-Cyber",
		"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 "mdomina/Qwen3-Coder-Next-Cyber" \
        --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": "mdomina/Qwen3-Coder-Next-Cyber",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mdomina/Qwen3-Coder-Next-Cyber with Docker Model Runner:
```
docker model run hf.co/mdomina/Qwen3-Coder-Next-Cyber
```

Qwen3-Coder-Next-Cyber

Cybersecurity domain specialization of Qwen/Qwen3-Coder-Next (the instruct model), built by Expert Upcycling + freeze and knowledge-injected via self-distillation — not by full continued-pretraining.

The 80B instruct backbone is frozen (instruction-following and coding ability preserved by construction); cybersecurity capability lives in 32 added experts per layer (512 to 544) that are the only trained part.

Research artifact. Downstream evaluation (cyber benchmarks) in progress.

What it is


Base	Qwen3-Coder-Next (instruct, 80B total / ~3B active, hybrid Gated DeltaNet + Attention + MoE, 512 experts top-10 + 1 shared, Apache-2.0)
Experts	512 to 544 (k=32 cyber experts added per layer, 48 layers)
Trainable	only {new experts 512-543, router, shared expert} = ~5.04B of ~85B; backbone frozen
Total params	~84.5B

How it was built

Expert upcycling: 32 new experts/layer cloned from high-utility experts (utility based on gradient-norm squared), router extended, top-k unchanged.
Warm-start: the 32 new experts were first continued-pretrained on a cyber corpus (~200M tokens), then their trained weights were transplanted onto the instruct backbone.
Self-distillation (this model): the model reads a domain passage in-context (teacher, open-book) and answers; the closed-book student (same weights) learns the answer. Trained only the 32 experts + router + shared (backbone + original 512 experts frozen), ~43k self-distilled QA, 1350 steps.

Why this recipe: naive SFT on new facts is unreliable and increases hallucination (Gekhman 2024); self-distillation from the model own open-book reading is in-distribution and data-efficient (arXiv 2412.14964). Knowledge lives in experts (frozen backbone means no forgetting), behavior in later adapters.

Knowledge domains (self-distillation sources)

MITRE ATT&CK techniques, OWASP (WSTG/Top-10), HackTricks, PayloadsAllTheThings, Atomic Red Team, adversary-emulation, OSCP/PEN-200, threat-intelligence, and detection tooling (osquery/Falco/Suricata). ~10.8k concepts, ~43k question-answer pairs.

Lineage note

An earlier artifact, mdomina/Qwen3-Coder-Next-Cyber-k32, applied the same upcycling but on the base (non-instruct) backbone via full CPT. This model corrects that: it uses the instruct backbone (so instruction-following is preserved) and injects knowledge via self-distillation instead of a large CPT.

Intended use & limitations

Domain-adapted assistant base for cybersecurity (offensive and defensive knowledge is symmetric). Research artifact; quantitative cyber-benchmark gains are not yet measured. Behavior beyond the base instruct template may need downstream SFT/RL.

Responsible use

Encodes offensive-security knowledge for authorized testing, defense, and education. Do not use against systems you do not own or lack permission to test.

License

Apache-2.0 (inherited from Qwen3-Coder-Next).

Downloads last month: -

Safetensors

Model size

85B params

Tensor type

BF16

Model tree for mdomina/Qwen3-Coder-Next-Cyber

Base model

Qwen/Qwen3-Coder-Next

Finetuned

(35)

this model