The H.AI cookbook provides code examples and guides to help developers use models developed by H Company.
| Version | Task | Notebook |
|---|---|---|
| Holo1.5 | Inference | Holo1.5 Quickstart |
| Holo1.5 | Inference | Holo1.5 Localization task via API |
| Holo1 | Deployment | Deploy Holo1 with vLLM on Nvidia |
| Holo1 | Inference | vLLM Navigation Task |
| Holo1 | Inference | vLLM Localization Task |
| Holo1 | Deployment | Deploy Holo1 on AWS SageMaker |
| Holo1 | Inference | SageMaker Navigation Task |
| Holo1 | Inference | SageMaker Localization Task |
H Company is pushing the boundaries of what our models are capable of achieving within a wide range of agentic scenarios. Our Holo1 model family, released in June 2025, provides a robust starting point for GUI agents, from which we’ve continued to improve the reliability and accuracy of our Action Vision Language Models (VLMs).
The Holo1.5 family breaks new ground by demonstrating state-of-the-art performance across benchmarks, establishing new baselines for all model sizes, from 3B to 72B. It excels in UI localization tasks such as Screenspot, Screenspot-V2, Screenspot-Pro, GroundUI-Web, Showdown and our own newly introduced WebClick benchmark. Holo1.5 also proves impressively adept in Content Understanding and Question Answering (QA) within web, computer and mobile use scenarios.
When used in the Surfer-H web agent system, Holo1.5 enables agents to navigate real applications with greater accuracy, reliability, and efficiency.
Holo1 is an Action Vision-Language Model (VLM) developed by H Company for use in the Surfer-H web agent system. It is designed to interact with web interfaces like a human user.
As part of a broader agentic architecture, Holo1 acts as a policy, localizer, or validator, helping the agent understand and act in digital environments.
Trained on a mix of open-access, synthetic, and self-generated data, Holo1 enables state-of-the-art (SOTA) performance on the WebVoyager benchmark, offering the best accuracy/cost tradeoff among current models. It also excels in UI localization tasks such as Screenspot, Screenspot-V2, Screenspot-Pro, GroundUI-Web, and our own newly introduced benchmark, WebClick.
Holo1 is optimized for both accuracy and cost-efficiency, making it a strong open-source alternative to existing VLMs.
- Holo1 3B - Qwen RESEARCH LICENSE
- Holo1 7B - Apache 2.0
- How to deploy Holo1 (3B or 7B) locally with vLLM on Nvidia
- Using OpenAI Client to invoke Holo1 for a navigation task
- Using OpenAI Client to invoke Holo1 for a localization task
- How to deploy Holo1 (3B or 7B) on Amazon SageMaker
- Using AWS SageMaker to Invoke Holo1 for a navigation task
- Using AWS SageMaker to Invoke Holo1 for a localization task
BibTeX:
@misc{andreux2025surferhmeetsholo1costefficient,
title={Surfer-H Meets Holo1: Cost-Efficient Web Agent Powered by Open Weights},
author={Mathieu Andreux and Breno Baldas Skuk and Hamza Benchekroun and Emilien Biré and Antoine Bonnet and Riaz Bordie and Matthias Brunel and Pierre-Louis Cedoz and Antoine Chassang and Mickaël Chen and Alexandra D. Constantinou and Antoine d'Andigné and Hubert de La Jonquière and Aurélien Delfosse and Ludovic Denoyer and Alexis Deprez and Augustin Derupti and Michael Eickenberg and Mathïs Federico and Charles Kantor and Xavier Koegler and Yann Labbé and Matthew C. H. Lee and Erwan Le Jumeau de Kergaradec and Amir Mahla and Avshalom Manevich and Adrien Maret and Charles Masson and Rafaël Maurin and Arturo Mena and Philippe Modard and Axel Moyal and Axel Nguyen Kerbel and Julien Revelle and Mats L. Richter and María Santos and Laurent Sifre and Maxime Theillard and Marc Thibault and Louis Thiry and Léo Tronchon and Nicolas Usunier and Tony Wu},
year={2025},
eprint={2506.02865},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2506.02865},
}
This library is licensed under the Apache 2.0 License. For more details, please take a look at the LICENSE file.