LangChain State of AI 2024 Report

This year's LangSmith users have shown a significant interest in open-source model adoption, with top providers such as Ollama and Groq gaining momentum, breaking into the top 5. The most used LLM provider remains OpenAI, but users are increasingly opting for more flexible deployment options like cloud deployment. Developers are building more dynamic applications, with AI agents gaining traction and becoming a key trend. Observability is needed regardless of what framework is used to build an LLM app, and interoperability is supported by LangSmith. Python still dominates, but JavaScript usage grows, driven by web-first applications. The average number of steps per trace has doubled over the past year, signaling more complex workflows, while the average number of LLM calls per trace remains relatively modest. Organizations are using LangSmith's evaluation capabilities to automate testing and generate user feedback loops to create more robust, reliable applications.