Arena.ai (formerly LMArena) has unveiled "Agent Mode," a new feature that lets AI agents autonomously carry out tasks in real environments, alongside "Agent Arena," a leaderboard that ranks them. The company released an execution environment in which AI autonomously handles planning, tool use, and multi-step processing from a single prompt, and feeds actual user sessions directly into model evaluation. Founding Engineer Matt Hova and Product Lead Ted Moran demonstrated the system handling deep research and complex bash operations (demo video).
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.