Instructions to use Junhauwong/Surge-Cognition-4x8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Junhauwong/Surge-Cognition-4x8B with Transformers:

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

pipe = pipeline("text-generation", model="Junhauwong/Surge-Cognition-4x8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Junhauwong/Surge-Cognition-4x8B")
model = AutoModelForCausalLM.from_pretrained("Junhauwong/Surge-Cognition-4x8B")
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

vLLM

How to use Junhauwong/Surge-Cognition-4x8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Junhauwong/Surge-Cognition-4x8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Junhauwong/Surge-Cognition-4x8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Junhauwong/Surge-Cognition-4x8B

SGLang

How to use Junhauwong/Surge-Cognition-4x8B 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 "Junhauwong/Surge-Cognition-4x8B" \
    --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": "Junhauwong/Surge-Cognition-4x8B",
		"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 "Junhauwong/Surge-Cognition-4x8B" \
        --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": "Junhauwong/Surge-Cognition-4x8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Junhauwong/Surge-Cognition-4x8B with Docker Model Runner:
```
docker model run hf.co/Junhauwong/Surge-Cognition-4x8B
```

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.

Surge-Cognition-4x8B

The Apex of Local Intelligence.

Surge-Cognition-4x8B is an experimental Mixture-of-Experts (MoE) language model focused on reasoning, coding, medical understanding, and general conversational intelligence.

The model dynamically routes prompts across expert pathways to preserve strong reasoning depth, coding performance, and contextual conversation quality while maintaining efficient active parameter usage.

🌑 Capabilities

Dynamic Expert Routing
Activates specialized reasoning pathways depending on prompt intent.
Coding & Technical Workflows
Strong performance in software engineering, scripting, and structured generation tasks.
Reasoning & Problem Solving
Maintains stable chain-of-thought style reasoning and commonsense inference.
Medical & Knowledge Retention
Preserves clinical terminology comprehension and biomedical contextual understanding.
Long-form Conversation
Coherent multi-turn responses with stable conversational flow.

📊 Benchmark Results

Benchmark	Score
MMLU	0.6509
HumanEval pass@1	0.561
GSM8K	0.7324
HellaSwag acc_norm	0.7858

🧠 Model Details

Attribute	Value
Architecture	Mixtral-style Mixture of Experts
Experts	4 Experts
Active Parameters	~8B
Total Parameters	~32B
Precision	BF16
Context Window	8K

⚠️ Notes

This is an experimental community-built MoE merge model intended for:

research,
local inference experimentation,
AI SaaS workflows,
coding assistants,
and agentic systems.

Benchmark scores are self-evaluated using LM Evaluation Harness and are not officially verified.

Downloads last month: 32

Safetensors

Model size

25B params

Tensor type

BF16

Evaluation results

accuracy on MMLU
self-reported

0.651
pass@1 on HumanEval
self-reported

0.561
acc_norm on HellaSwag
self-reported

0.786
exact_match on GSM8K
self-reported

0.732