Published
LLM Multi-Round Router
LLM-based agentic router that uses language understanding to route complex agentic tasks.
Overview
The LLM Multi-Round Router uses an LLM itself to analyze incoming tasks and make routing decisions, providing the highest level of flexibility for agentic workflows.
How It Works
A meta-LLM (a language model used for routing decisions) analyzes each task description, understands its requirements, breaks it into steps, and assigns each step to the most capable model. The meta-LLM can also reassess routing decisions mid-task based on intermediate results.
Strategy
Uses LLM itself to route complex agentic tasks.
API Endpoint
autoroute:llmmultiroundrouter
Use Cases
- Advanced agent systems requiring dynamic routing
- When you need meta-routing capabilities
- Tasks that benefit from language understanding during routing
Best Practices
Meta-LLM Selection
The meta-LLM should be capable but not necessarily the most expensive model. A mid-size model with strong reasoning (e.g., GPT-4o-mini or Claude 3 Haiku) often provides the best routing quality-to-cost ratio.
Related Models
- KNN Multi-Round Router — For similarity-based agentic routing
- Causal LLM Router — For single-round language-based routing