LMArena (Chatbot Arena)

Completely Free

The gold standard in blind LLM evaluation: try anonymous models on the same prompt, vote for the best, and reveal true Bradley-Terry Elo rankings.

Editorial note

UC Berkeley LMSYS project. Test mystery canary models before official public launch. See our step-by-step masterclass on /compare.

Details

Metadata is curated manually; anything we could not verify is left out rather than guessed.

comparebenchmarkmodelseloblind-testing

Ad-free AI answer engine that summarises pages and never tracks you.

Anonymous chat with several models through DuckDuckGo's privacy proxy.

Chat interface running open-weight models hosted by Hugging Face.

Multi-LLM consensus platform: Ask one prompt and compare answers from Claude, ChatGPT, Gemini, and open models in parallel.

Browser workspace for prompting Gemini models and testing an API key.

Very fast hosted inference for open models, with a free developer tier.