d-s-b/Router
TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.3BQuant:BF16Context Size:32kPublished:Aug 17, 2025Architecture:Transformer0.0K Featherless Exclusive Cold
The d-s-b/Router is a compact 0.3 billion parameter model, fine-tuned to act as a reasoning router for queries. It classifies input queries into 'no_reasoning', 'low_reasoning', or 'high_reasoning' categories, making it ideal for directing tasks to specialized LLMs based on complexity. This model excels at efficiently triaging diverse reasoning tasks, leveraging its 32768 token context length for effective classification.
Loading preview...
Router Model Overview
The d-s-b/Router is a specialized language model designed to classify input queries based on their reasoning complexity. This 0.3 billion parameter model, fine-tuned from the Gemma 3 270M base, serves as an efficient routing mechanism for diverse reasoning tasks.
Key Capabilities
- Reasoning Classification: Accurately categorizes queries into three distinct reasoning levels:
no_reasoning: For direct factual lookups or simple recall (e.g., "What is the capital of France?").low_reasoning: For tasks requiring basic arithmetic, comparisons, or single logical steps (e.g., "If John has 5 apples and eats 2, how many are left?").high_reasoning: For complex problems demanding multi-step logic or deep problem-solving (e.g., "Prove that the sum of two even numbers is always even").
- Efficient Task Routing: Enables intelligent distribution of queries to appropriate downstream models or systems based on their identified reasoning requirements.
Good for
- Orchestrating LLM Workflows: Ideal for developers building systems that need to route user queries to different specialized LLMs (e.g., a factual QA model, a mathematical solver, or a complex reasoning engine).
- Optimizing Resource Usage: By pre-classifying queries, it helps ensure that computationally intensive models are only engaged for high-reasoning tasks, improving overall efficiency.
- Building Intelligent Agents: Useful for creating agents that can dynamically adapt their approach based on the complexity of the user's request.