Description
Browser-operating web agents mostly rely on closed models researchers can't reproduce or improve. MolmoWeb is an open multimodal web agent built by Ai2 that clicks, types, scrolls and navigates to complete natural-language tasks.
The repo has the agent, inference client, benchmarks and everything needed to train, with open models and data.
Autonomous:Click, type and navigate.
Multimodal:Reads screenshots.
Benchmarks:Reproducible.
Open:Models and data.
The repo has the agent, inference client, benchmarks and everything needed to train, with open models and data.
Features
Autonomous:Click, type and navigate.
Multimodal:Reads screenshots.
Benchmarks:Reproducible.
Open:Models and data.
