Meru

At Meru, we help developers build custom, task-specific LLMs that work just as well as ChatGPT, but run up to 40x faster, run on commodity CPU hardware, and can be deployed on premise. As developers, we built applications on top of OpenAI and got frustrated because they were so slow. Our latency was high because we used a foundational model that is generalized, and designed to solve a wide variety of problems. It turns out when you are trying to solve a specific problem, you can train a much smaller model that will perform just as well, but run up to 40x faster and on commodity hardware. Today, developers everywhere are replacing OpenAI’s generic endpoints with task-specific, highly responsive endpoints built on Meru.