Source item

We launched a killer feature on @cfosilvia last night. Users now have access to a state-of-the-art AI model router that predicts which AI model is best suited to answer a specific query and then quickly retrieves the relevant information from the necessary model. Example: we don’t need to use the latest Opus model to answer “what time is it?” or “what is Apple’s stock price?” We can route these simple queries to cheaper models that are actually tuned to better answer these types of questions. I believe model routing will become a core component of every AI product because leaders want performance with minimal costs. The latest models are breathtakingly good, but they are also too expensive for queries that don’t need that level of power. An added benefit we saw during testing of the model router is that latency decreased for many queries too. This means the model router decreased costs and increased speed. Win, win. The engineering team at Silvia is elite in my opinion. They are out-innovating some of the top AI model labs in the world. We will continue sharing updates on various features or innovations as we build $BRR and Silvia.