Perciqa/Aurora-Code-Mini-V1

TEXT GENERATIONConcurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 30, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Aurora-Code-Mini-V1 is a 14.8 billion parameter dense coding model developed by Perciqa, fine-tuned from Qwen3-14B. It specializes in agentic coding tasks, including code generation, debugging, review, and refactoring across 40+ languages. Designed for on-premise deployment, it offers high-quality coding assistance without cloud dependencies, requiring approximately 28 GB VRAM at BF16 or 8 GB with 4-bit quantization.

Loading preview...

Aurora-Code-Mini-V1: Compact, Capable, Canadian Coding Model

Aurora-Code-Mini-V1 is a 14.8 billion parameter dense coding model developed by Perciqa, a Canadian AI company. It is fine-tuned from the Qwen3-14B base model using a proprietary dataset of agentic coding instruction pairs. This model is designed for developers seeking fast, high-quality coding assistance that can be deployed and controlled on their own infrastructure, ensuring data privacy and avoiding cloud dependencies.

Key Capabilities

  • Comprehensive Code Assistance: Excels in code generation, debugging, refactoring, and test writing across over 40 programming languages.
  • Agentic Tasks: Supports multi-step tool use, planning, and repository-level reasoning.
  • Code Quality & Security: Capable of identifying security issues, suggesting refactors, and explaining design trade-offs during code review.
  • On-Premise Deployment: Optimized for deployment on local infrastructure, requiring approximately 28 GB VRAM for BF16 or 8 GB for 4-bit quantization.

Training Approach

The model was fine-tuned using a rigorous, multi-stage approach involving proprietary data curation and Parameter-Efficient Fine-Tuning (LoRA). This method preserves the base model's general reasoning while specializing in developer-centric tasks, with extensive evaluation to ensure high token accuracy and generalization.

Recommended Use Cases

  • Secure Development Environments: Ideal for organizations requiring strict data governance and on-premise AI solutions.
  • Automated Code Workflows: Suitable for integrating into CI/CD pipelines for automated code generation, review, and testing.
  • Developer Tooling: Enhancing IDEs and developer platforms with intelligent coding assistance.