LiveBench

LiveBench is an objective, contamination-free benchmarking platform for evaluating large language models (LLMs) across 23 tasks in 7 categories, including reasoning, coding, mathematics, data analysis, language, and instruction following. Updated biannually, it enables users to view and compare model scores, costs per successful task, and subtask performance, with sortable leaderboards and analytical cost/quality views. LiveBench is designed for organizations, researchers, and product teams seeking transparent, up-to-date performance and cost benchmarking for LLMs.