A promise of a next generation of AI first fulfilled by H

MUMBAI, India, Nov. 21, 2024 -- H, the leader in agentic AI, proudly announces the launch of its groundbreaking product, Runner H, a first cloud-based web agent. With the beta release of Runner H 0.1, H is breaking through the limitations of plateauing scaling laws, taking a significant step toward artificial superintelligence (ASI).

Put AI to work for you with Runner H

"We are thrilled to bring agentic AI to market with Runner H," said Charles Kantor, CEO of H. "This is about empowering people to harness AI like never before. Runner H allows users to automate workflows, streamline tasks, and own the web"

Agentic AI is revolutionizing how we approach repetitive and manual tasks. H leverages its proprietary foundation and action models to deliver advanced agents capable of reasoning, planning, and execution. These agents automate complex, multi-step, repetitive workflows while remaining under human oversight, making them indispensable tools for businesses and individuals alike.

"Runner H is more than just a productivity tool," Kantor added. "It's about empowerment—helping people focus on what matters most by automating the mundane. It frees time for creativity, innovation, and meaningful human connections."

Designed for versatility, Runner H can seamlessly integrate with multi-system and legacy solutions in large enterprises while also offering entrepreneurs a virtual team of agents. Whether handling routine tasks or executing intricate QA processes, Runner H is paving the way for a new era of human-machine collaboration.

What Can Runner H Do?

Runner H is addressing the disruption, vertical by vertical, of billion dollars markets such as :

Robotic Process Automation (RPA): Automating Hiring Processes with Ease

Runner H revolutionizes hiring, enabling companies to automate their entire recruitment process with a single prompt. From posting job descriptions to aggregating resumes and sending tailored follow-ups, Runner H eliminates manual effort. It enriches applicant data with insights from LinkedIn and other platforms, updates CRMs automatically, and even handles onboarding. With Runner H, businesses can cut weeks of work into moments, saving time and resources while finding the perfect candidate faster than ever.

Quality Assurance (QA): Reimagining Entire e-commerce funnel testing

Testing websites and applications can be time-consuming and resource-intensive, but Runner H changes the game. It seamlessly adapts to user interface changes, automates repetitive testing processes, and navigates complex workflows with ease – whether for validating page availability, simulating real user actions, or ensuring compatibility across payment methods. By reducing maintenance burdens, it frees developers to focus on innovation and creativity, driving value where it matters most.

Business Process Outsourcing (BPO): Taking Control of Billing Workflows

Inefficiencies in billing workflows - e.g. in dental insurance - requires time and money, but not anymore. Runner H can empower businesses to reclaim these workflows by automating the most tedious tasks, e.g. fetching insurance plans, compiling notes, analyzing and submitting claims, and tracking reimbursements—all with precision. By streamlining these processes, Runner H eliminates outsourcing dependencies, reduces delays, and helps businesses take control of their financial operations.

The Competitive Edge of Runner H 0.1

The agent Runner H 0.1 delivers state-of-the-art performance and outperforms Anthropic's Computer Use by 29% in the public benchmark WebVoyager. Underlying this technology, H's compact models (LLM – Runner's brain, and VLM – Runner's eye, both 2B parameters) are also surpassing larger alternatives from competitors like Mistral and Meta. This success demonstrates the potential of smaller, more focused models to outperform generalist models at a fraction of the cost.

As shown below, H's two-billion-parameter model performed extremely well on an average of code and function calling, outperforming much bigger models. H-LLM is the backbone of H-VLM and can also be used in H agents for text-only roles.

H trained and specialized 3B parameters VLM to perceive, understand, and interact with graphical and text elements. H-VLM is by far the strongest small model in localization based on the Screenspot's benchmark (for graphical user interface actions), outperforming the likes of GPT4o (OpenAI), Sonnet 3.5 (Anthropic), Pixtral Large (Mistral), Llama (Meta). Impressively, H model is much more accurate than the very large generalist models, while being orders of magnitude cheaper and faster to serve.

By focusing on the web as its initial environment, Runner H delivers immediate value in areas like web testing, process automation, and outsourcing. The ultimate vision is universal automation : AI agents capable of navigating any graphical interface.

"Today's large language models are just one piece of the puzzle," said Kantor. "With Runner H, we are opening a new era in AI : one that takes action in the real-world. Agentic AI isn't just fascinating, it is truly transformative for the Human-Machine interface"

Learn more about the vision behind Runner H : https://www.hcompany.ai

Private beta waitlist is now open

H is inviting developers to join the private beta for Runner H 0.1, offering tools to create scalable, production-ready automation using natural language instructions or visual demonstrations.

The beta includes:

• the API to call off-the-shelf and managed agents running in the cloud
• the H-Studio to review and edit past and live runs

Developers can sign up for the waitlist for the private beta at https://www.hcompany.ai/waitlist

This is the first step in H's journey to democratize AI Runners.

"Runner H is just the beginning," said Kantor. "We are excited to start with the web and expand to other environments. The future of automation is limitless, and we are proud to lead the charge. Looking forward to the future where people will put products like Runner H to automate for them"

About Agentic AI:

Agentic AI is designed to act autonomously, planning actions and making decisions to achieve specific user-defined goals. Unlike traditional AI models, which require explicit inputs and produce predetermined results, agentic AI can develop plans and operate independently, using external tools and resources to accomplish complex tasks. To build the best agents, H's agents rely on specialist models that are exceptionally brilliant in a few domains. H's in-house VLMs, initiated from LLMs of its own, focus on UI capabilities, starting with understanding and localizing the user interface. 

About The H Company (H)

H is a foundation model and agentic AI company, founded in late 2023 by Charles Kantor (CEO) and Laurent Sifre (CTO), one of DeepMind's leading scientists. In May 2024, H announced an initial fundraising round of $220 million from global firms and investors including Accel, Eric Schmidt, Yuri Milner, Xavier Niel, Bernard Arnault, Amazon, UiPath Inc, Samsung, Bpifrance, FirstMark, Elaia Partners, Eurazeo. https://www.hcompany.ai/

Logo: https://mma.prnewswire.com/media/2564513/The_H_Company_Logo.jpg