Salesforce/xRouter

Salesforce AI Research's xRouter is a 7.6 billion parameter instruction-tuned language model with tool-calling capabilities, built upon Qwen/Qwen2.5-7B-Instruct. This system uses reinforcement learning (DAPO) with cost-aware reward shaping to dynamically select optimal LLMs from over 20 available models. It excels at orchestrating multi-model responses for complex reasoning tasks while significantly reducing operational costs and maintaining performance.