tools
LMArena
LMArena, now Arena, lets people compare AI model responses, vote on them, and explore public leaderboards.
Arena is a web platform for testing and comparing third-party AI models. Users can run anonymous model battles, chat directly with models, compare responses side by side, vote, and explore leaderboards based on community feedback.
It also includes multimodal tools and Agent Mode for research, coding, file work, web search, image generation, and sandboxed tasks. The public service is free, while commercial AI evaluation services are available separately; no official pricing page was found.
Features
- Compare anonymous AI model responses in side-by-side battles
- Vote on responses to shape public model leaderboards
- Use Battle, Direct, and Side-by-Side chat modes
- Run Agent Mode with web search, uploads, coding, and bash
- Generate and evaluate images, videos, and other media
- Connect GitHub repositories, review diffs, commit, push, and open pull requests
- Access open datasets, research papers, and evaluation methodology
Use cases
- Compare models before choosing one for a writing or coding task
- Evaluate model responses through anonymous community voting
- Build and modify applications with Agent Mode and GitHub
- Run research workflows using web search, files, and a sandbox
- Inspect leaderboards for text, vision, coding, search, image, and video tasks