H Company launches Holo2-235B-A22B for UI localization
h company released holo2-235b-a22b preview, a 235b parameter model for ui element localization. it hits 78.5% on screenspot-pro and 79.0% on osworld g, setting a new sota record. the model uses agentic localization to iteratively refine predictions, improving accuracy by 10-20%.
H Company released Holo2-235B-A22B Preview, a 235 billion parameter model designed for UI element localization, achieving a new State-of-the-Art (SOTA) record of 78.5% on ScreenSpot-Pro and 79.0% on OSWorld G. The model utilizes 'agentic localization' to iteratively refine predictions on high-resolution 4K interfaces, yielding a 10-20% relative accuracy improvement across all model sizes compared to single-step predictions. Trained using SkyPilot to coordinate workloads across multiple cloud providers, the model is available now on Hugging Face.