The benchmark tests the entire "agent" system – not just the model, but also the software scaffolding around it that handles prompts, parses outputs, and manages the interaction loop.

→ View original post on X — @alexalbert__,