smashingtags/nova-router-1.5b

TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 10, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

smashingtags/nova-router-1.5b is a 1.5 billion parameter tool-routing model developed by smashingtags, fine-tuned from Qwen2.5-Coder-1.5B. It is specifically designed for the Eight.ly OS to accurately select NAS management tools based on user requests, achieving 99% tool-selection accuracy on the Eight.ly evaluation suite. This model functions as a hidden dependency, routing requests to appropriate tools across Docker, storage, VMs, LXC, file sharing, and system administration, with a context length of 32768 tokens.

Loading preview...

Nova Router 1.5B Overview

Nova Router 1.5B is a specialized 1.5 billion parameter tool-routing model developed by smashingtags, built upon the Qwen2.5-Coder-1.5B architecture. Its primary function is to act as an intelligent router within the Eight.ly OS, determining which NAS management tool to invoke in response to a user's request. This model is a critical, hidden component of the Nova local AI assistant, working in conjunction with a conversational 'talker' model.

Key Capabilities & Features

  • Tool Routing: Accurately selects from 41 NAS management tools across 6 domains (Docker, storage, VMs, LXC, file sharing, system administration).
  • High Accuracy: Achieves 99% tool-selection accuracy on the Eight.ly evaluation suite, with zero false positives for queries not requiring a tool.
  • Integrated Design: Operates as a 'router-in-front' system, where it processes user queries, performs intent classification and retrieval, and then routes to the appropriate tool before a separate 'talker' model generates the final reply.
  • Optimized for Eight.ly OS: Specifically trained on the Eight.ly tool-calling dataset to seamlessly integrate and enhance the functionality of the Eight.ly local AI assistant.
  • Efficient Deployment: Available in quantized GGUF formats (Q8_0 at ~1.6 GB, Q4_K_M at ~1.0 GB) for efficient local deployment, automatically installed with any Nova talker model.

When to Use This Model

This model is ideal for applications requiring precise and efficient tool selection, particularly within NAS management or similar environments where a dedicated routing layer is beneficial. Its high accuracy and specialized training make it suitable for integrating complex tool-use capabilities into AI assistants, ensuring the correct tool is invoked for specific user intents.

Finetunes
1 models