A benchmark where LLMs make memes from current news

Memebench is a public benchmark platform that evaluates how well different language models generate memes from current news. Users browse AI-created memes, vote on their quality and humour, and see how various models rank on a live leaderboard. The platform measures something often overlooked in AI evaluation: how well models understand topical context, cultural references, and comedic timing. It's particularly useful for researchers studying generative AI capabilities, AI engineers testing meme generation models, and anyone curious about how different language models approach creative and contextual tasks. By crowdsourcing meme quality judgments rather than relying on automated metrics, Memebench provides real human feedback on model performance. The leaderboard updates regularly as new memes are generated from trending news, creating a dynamic benchmark that reflects both model capability and current events.