LLMBoard.ai

LLMBoard.ai answers a question that changes every few weeks: which model should we actually use? It maintains a continuously updated catalogue of model records and benchmark results, then presents them as leaderboards organised by task — overall chat and LLM, agentic work, coding, reasoning, math, knowledge, instruction following, plus separate tracks for vision, image generation and editing, video, text-to-speech, speech-to-text and embeddings. Each ranking is backed by named benchmarks such as GPQA, MMLU-Pro, AIME, SWE-Bench Verified, LiveCodeBench, MATH, HumanEval and MMMU-Pro.

What distinguishes it from a single-number ranking is that it deliberately separates signals. Capability scores sit apart from price, runtime latency and provider reliability, so a model that tops a benchmark is not automatically presented as the best value or the most dependable at scale. There are dedicated pricing pages for chat tokens, images, video and audio, plus speed and reliability views. A scoring and data section documents how models are mapped and scored, and each benchmark page shows raw model order and coverage so you can judge the evidence behind a position.

The site is aimed at engineering teams, product managers and researchers evaluating models for a specific workload, as well as anyone tracking how open models compare with frontier proprietary ones. It is an information and comparison resource — it does not host the models themselves, and results should be treated as a starting point for your own testing.