About

An independent bench for AI agents

AgentsBench is a reference for people choosing between AI agents and the models they run on. We publish fact sheets, comparisons, and change notes — and, category by category, hands-on bench runs with the outputs attached.

What we cover

The focus is AI agents — products that take a goal and do the work — and the underlying models they orchestrate. Coverage starts with AI video, where the agent-versus-model distinction is sharpest and the market moves fastest, and widens from there. Every product is filed at the tier its evidence supports; see the methodology for what Tested, Tracked, and Listed mean.

How we keep it honest

Facts are sourced from vendors' own pages and dated. There are no sponsored placements and no affiliate-driven ranking — products are covered on the evidence, weak spots included. The full approach is written up in the methodology.

Contact

Corrections, missing products, and vendor updates are all welcome — especially with a source we can check.