Arena
Platform where you compare two anonymous AI models on the same prompt, and your votes feed public leaderboards.
What it does
In battle mode, Arena (formerly LMArena) sends your prompt to two anonymous models. You compare the answers side by side, vote, and only then learn which models they were. Alternatively, you chat directly with a model you pick. From the votes, Arena calculates public leaderboards, for example for text, WebDev, vision, documents, search, and image and video generation.
Who it suits
Anyone who wants to check how current models handle their own tasks before committing to one, such as writing, code or website prototypes. Model providers and businesses can also commission paid evaluations based on human ratings.
Data protection for businesses
The operator is Arena Intelligence, Inc. in San Francisco; data may be transferred to the USA and other countries. The vendor does not offer EU data hosting. Your inputs go to the respective model providers, which are not required to keep them confidential. De-identified conversations may end up in public research datasets and be passed to model providers to improve their models; this cannot be switched off. The vendor does not mention a data processing agreement, and there is no business plan for using the service. Keep confidential or personal data out of it.
Limits
You must be at least 18 to use it. The terms of use prohibit commercially exploiting the service. The vendor does not mention an API for your own applications. The leaderboards reflect the preferences of those who vote and do not replace tests with your own data.