Instructions to use Perciqa/Aurora-Code-Mini-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Perciqa/Aurora-Code-Mini-V1 with Transformers:

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

pipe = pipeline("text-generation", model="Perciqa/Aurora-Code-Mini-V1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Perciqa/Aurora-Code-Mini-V1")
model = AutoModelForCausalLM.from_pretrained("Perciqa/Aurora-Code-Mini-V1", 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 Perciqa/Aurora-Code-Mini-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Perciqa/Aurora-Code-Mini-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Perciqa/Aurora-Code-Mini-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Perciqa/Aurora-Code-Mini-V1

SGLang

How to use Perciqa/Aurora-Code-Mini-V1 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 "Perciqa/Aurora-Code-Mini-V1" \
    --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": "Perciqa/Aurora-Code-Mini-V1",
		"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 "Perciqa/Aurora-Code-Mini-V1" \
        --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": "Perciqa/Aurora-Code-Mini-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Perciqa/Aurora-Code-Mini-V1 with Docker Model Runner:
```
docker model run hf.co/Perciqa/Aurora-Code-Mini-V1
```

Aurora-Code-Mini-V1

Compact. Capable. Canadian.

Aurora-Code-Mini-V1 is a 14.8B dense coding model built by Perciqa, a Canadian AI company. Fine-tuned from Qwen3-14B on a highly curated, proprietary dataset of agentic coding instruction pairs, Aurora-Code-Mini-V1 is designed for developers who need fast, high-quality coding assistance — without cloud dependencies, usage limits, or black boxes.

License: Apache 2.0
Hardware: Requires ~28 GB VRAM at BF16, or ~8 GB with 4-bit quantization.
Made in Canada 🇨🇦

What Aurora-Code-Mini-V1 Does

Aurora-Code-Mini-V1 is tuned specifically for developers who need a model they can deploy, audit, and fully control on their own infrastructure.

Code Generation: Write functions, classes, and complete programs across 40+ languages.
Debugging: Identify root causes and produce clear, actionable fixes.
Code Review: Flag security issues, suggest refactors, and explain tradeoffs.
Agentic Tasks: Multi-step tool use, planning, and repository-level reasoning.
Refactoring: Modernize legacy code, apply design patterns, and improve maintainability.
Test Writing: Generate unit tests, integration tests, and comprehensive test suites.

No black boxes. No data leaving your infrastructure. Your model, your terms.

Quickstart

Install

pip install "transformers>=4.51.0" accelerate peft

Transformers (Adapter)

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

model_name = "Qwen/Qwen3-14B"
adapter_name = "Perciqa/Aurora-Code-Mini-V1"

tokenizer = AutoTokenizer.from_pretrained(model_name)
base = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto",
)
model = PeftModel.from_pretrained(base, adapter_name)

system_prompt = (
    "You are Aurora, an AI code assistant built by Perciqa. "
    "You help developers write, review, and understand code. "
    "You provide clear, correct, and complete solutions. "
    "When you're unsure, you say so."
)

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "Write a Python function to merge two sorted lists."},
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=1024,
    temperature=0.7,
    do_sample=True,
)
response = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)
print(response)

vLLM (Recommended for Production)

pip install vllm
vllm serve Perciqa/Aurora-Code-Mini-V1 --max-model-len 32768

Query via the OpenAI-compatible API:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="token-abc123")

response = client.chat.completions.create(
    model="Perciqa/Aurora-Code-Mini-V1",
    messages=[
        {"role": "system", "content": "You are Aurora, an AI code assistant built by Perciqa."},
        {"role": "user", "content": "Refactor this function to be more Pythonic."},
    ],
    max_tokens=1024,
)
print(response.choices[0].message.content)

Ollama

ollama run hf.co/Perciqa/Aurora-Code-Mini-V1

Training Approach

(Note: Specific dataset metrics, teacher model names, and internal training configurations are kept proprietary to protect Perciqa's intellectual property.)

Aurora-Code-Mini-V1 is fine-tuned from the Qwen3-14B base model using a rigorous, multi-stage approach:

Proprietary Curation: Trained on a carefully curated, high-quality dataset of agentic coding instruction pairs spanning critical developer workflows, including generation, debugging, refactoring, and testing.
Parameter-Efficient Fine-Tuning: Optimized using Low-Rank Adaptation (LoRA) to preserve the base model's robust general reasoning capabilities while specializing in high-fidelity, developer-centric tasks.
Quality Assurance: Checkpoints were extensively evaluated on held-out validation sets to optimize for low loss, high token accuracy, and strong generalization without overfitting.

Model Details

Field	Value
Architecture	Dense Transformer (GQA)
Total Parameters	14.8B
Transformer Layers	40
Attention Heads	40 (Q) / 8 (KV)
Context Length	131,072 tokens (native)
Base Model	Qwen3-14B
License	Apache 2.0
Hardware (BF16)	~28 GB VRAM
Hardware (4-bit)	~8 GB VRAM

System Prompt

For optimal performance, we recommend using the following system prompt:

You are Aurora, an AI code assistant built by Perciqa.
You help developers write, review, and understand code.
You provide clear, correct, and complete solutions.
When you're unsure, you say so.

About Perciqa

Perciqa is a Canadian AI company building enterprise models and tools that organisations can deploy, audit, and fully control — on their own infrastructure, on their own terms. Founded in 2023 and based in Canada 🇨🇦.

perciqa.com · GitHub

License

Aurora-Code-Mini-V1 is released under the Apache 2.0 License.

Made with ♥ by Perciqa 🇨🇦

Downloads last month: -

Safetensors

Model size

15B params

Tensor type

BF16

Model tree for Perciqa/Aurora-Code-Mini-V1

Base model

Qwen/Qwen3-14B-Base

Finetuned

Qwen/Qwen3-14B

Adapter

(541)

this model