Introducing the Arena Alignment Index, our new benchmark measuring safety and alignment of AI agents in real-world use.
Built from 90K+ real-world agent sessions across 27 models, the index measures three critical signals:
- Unauthorized Action (UA): Taking actions beyond the
Last call: proposals for the Fall 2026 cycle of Arena’s Academic Partnerships Program are due tonight at midnight.
Up to $50,000 per project is available for tenure-track faculty at U.S. universities researching the scientific foundations of AI evaluation: evaluation and ranking
Reminder: the Fall 2026 cycle of Arena's Academic Partnerships Program has begun!
We're accepting new proposals from tenure-track faculty at U.S. universities working on the scientific foundations of AI evaluation: evaluation and ranking methodology, measurement and validity,
This Week in the Arena: Gemini is back, and Terra got even more efficient.
- Gemini 4 Carbon (High) by @GoogleDeepMind landed #1 in Text Arena, #2 in Agent Arena, and #4 in Code Arena: WebDev. It reshaped the Pareto frontier across all three arenas. See first impressions from
Your agent would cost 260% more without the cache.
We analyzed +26K Agent Arena sessions from September 12 to October 23, 2026, spanning 30 agents and 13 labs.
Caching dramatically changes the economics of a typical task. If every cached token were instead billed at the model's
