AutoMix
Automatic model mixing — combines outputs from multiple LLMs based on query complexity.
AutoMixmixingsingle-roundroutingensemble
Automatic model mixing — combines outputs from multiple LLMs based on query complexity.
Causal Language Model router — leverages language understanding to predict the best LLM for each query.
Elo Rating based routing — competitive, dynamic routing that adapts to changing LLM performance.
Graph-based personalized router that learns user preferences for tailored LLM selection.
Graph-based routing — models query-LLM relationships as a graph for sophisticated routing decisions.
Hybrid LLM routing strategy — combines multiple routing approaches for robust, production-ready performance.
KNN-based agentic router for complex, multi-step agent tasks requiring task decomposition.
K-Nearest Neighbors based routing — simple and effective query routing using similarity to training examples.
Baseline strategy that always routes to the largest model — maximum quality when cost is not a concern.
LLM-based agentic router that uses language understanding to route complex agentic tasks.
Matrix Factorization based routing — predicts best LLM by decomposing the query-LLM interaction matrix.
Multi-Layer Perceptron based routing — neural network-powered routing with learnable parameters.
GNN-based personalized router that uses user features to tailor LLM selection.
Pre-trained Router-R1 model for multi-turn conversations with context-aware routing.
Dual Contrastive learning based routing — learns query-LLM compatibility with limited labeled data.
Baseline strategy that always routes to the smallest available model — maximum cost efficiency and minimal latency.
Support Vector Machine based routing — leverages learned decision boundaries for high-dimensional query spaces.
Learn how to route requests to the best LLM models based on your needs, budget, and performance requirements.