MicroLLM Lab – Try 7 tiny LLM's in the browser (stateofutopia.com)

MicroLLM Lab has launched a user-friendly platform allowing users to test seven compact language models (LLMs) directly in their web browsers. Each 135M model is evaluated on objective metrics such as runtime, accuracy, and speed, which are measured in tokens per second. Users can choose models and execute tests to gauge their performance, with results displayed via charts that illustrate the latest performance metrics for each model without any data leaving the user's machine. This initiative is significant for the AI/ML community as it democratizes access to LLM benchmarking, enabling developers and researchers to easily compare different models' performance. The platform generates verifiable performance certificates detailing device hardware, peak and sustained token throughput, encouraging transparency and sharing within the community. By providing a JavaScript-based benchmarking tool that facilitates real-time model evaluation, MicroLLM Lab streamlines the process of understanding model strengths and weaknesses, fostering improved model development and innovation.