Omnionix-AI/avara-x1-mini

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Mar 9, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Avara X1 Mini is a 1.5 billion parameter language model developed by Omnionix, based on the Qwen2.5 architecture with a 32768 token context length. It is fine-tuned for technical reasoning, excelling in code, mathematics, and logical problem-solving. The model is designed to provide a grounded and supportive AI personality, making it suitable for applications requiring precise technical understanding combined with helpful interaction.

Loading preview...

Avara X1 Mini: Technical Reasoning with a Supportive Persona

Avara X1 Mini, developed by Omnionix, is a 1.5 billion parameter model built upon the Qwen2.5 architecture. It is specifically fine-tuned to balance strong technical reasoning capabilities with a grounded and supportive conversational identity. The model utilizes the ChatML format and incorporates native Omnionix system logic for its identity.

Key Capabilities & Training Focus

This model's training methodology, leveraging the Unsloth library, focused on maximizing reasoning performance within a compact footprint. Its high-density dataset blend includes:

  • Code: Extensive training on The Stack (BigCode) for professional-grade programming logic.
  • Mathematics: Specialized datasets for step-by-step mathematical problem-solving and competition-level math.
  • Logic: Integration of Open-Platypus for enhanced deductive reasoning and precise instruction following.

Ideal Use Cases

Avara X1 Mini is well-suited for applications requiring:

  • Technical Assistance: Providing accurate and logical responses to coding, mathematical, or general technical queries.
  • Instruction Following: Executing complex instructions with high fidelity due to its logic training.
  • Supportive AI Interactions: Scenarios where a helpful and grounded AI persona is beneficial alongside strong reasoning.

For developers, a LoRA adapter and a Q4_K_M GGUF version are also available.