100% task success across all demos 💸 ~50% lower token usage per step 🧠Works with small local models (3B–7B) Vision-only agents fail systematically on the same tasks ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results