route llm intelligently routes users query to a relevant Lodge language model to get a better response optimizing the quality optimizing the cost so route llm based on users query it optimizes for latency optimizes for cost and optimizes for Quality so without route llm it's going to increase the cost increase in latency and low quality for example a simple question like what is weather what is a large language model so these all simple questions can be answered by weak llm and that is going to cost little compared to a strong llm but for complex query a strong llm will be used to give us more accurate answer so route LM combines coding ability of Sona 3. 5 reasoning ability of 01 preview hard coding ability of 01 mini 1 million context of Gemini and the speed of 40 mini we have all these features in chat llm for teams this is mainly focused on user who are interested in using rout llm in a user interface rather than hardcoding it yourself so when you ask a simple question what is llm now you can see it's routed to GPD 40 mini because this is a basic question and a small or a weak model is able to answer the question for you as a result it optimizes in cost now if I have any complex question write a pytorch code to developer llm now this is routed to Cloe Sona 3. 5 as you can see here it's writing the code for you in pytorch now it's giving us the steps to run the code and create a large language model thanks to aboc for sponsoring this video you have the route llm feature in chat llm for teams which you can start today with $10 per user per month and I'll put the link in the description for you to sign up and this platform has multiple models for you to try it has various features such as access to large language model user management and Reporting chat with large PDF and create custom chatbot we're going to try various queries and see how this is routing to different models that's exactly what we're going to see today let's get [Music] started so now we are going to sign up by clicking this get started after login you have an option here at the top so click on that then you have option to choose the model you want to use but in my case I'm going to use route llm just click on that and now we are ready to test and see how it's going to perform based on our question I'm going to ask create a snake game in Python now it's routing to clo son 3.
5 and now it's writing the code for me Cloe is better in regards to programming so we can clearly see that this model is automatically redirecting the question to clo Sonet model I'm going to provide a logical and reasoning question and then let's see what model is being used so here you can see it's routed to 01 mini considering it's one of the top performing logical and reasoning model and it's thinking now now based on the question you can see how much complex the answer is so the model was able to go through that and give us the answer for the question we also have multiple option regenerating using GPT 40 clot 3. 5 40 Min so let me try 40 so now it's regenerating the same answer using the 40 model but this is without reasoning ability because o1 model will have reasoning ability by default I'm going to provide a rag question what are the latest advancements in Vector database the relevant data with the different chunks I'm providing those chunks and let's see what this model is going to do I'm going to say use below data to answer about question so for rack Pros as you can see it's using GPT 40 mini which is enough to answer the question based on the context now I uploaded PDF and ask it to explain at that time it is routing to GPT 40 and it's giving me key points from the uploaded PDF rather than us manually choosing the model and giving them the task using route Helm simplifies the process of automatically choosing the right model for your task so overall rout LM combines the coding ability ility for Sona 3.
