| November 21, 2024
H, the Paris-based AI startup founded by Google alums, made headlines last summer when it announced a staggering $220 million seed round before releasing a single product. However, the company faced turbulence just three months later when three of its five co-founders left, citing “operational and business disagreements.”
Despite these challenges, H has introduced its first product: Runner H, an “agentic” AI designed for businesses and developers tackling tasks like quality assurance and process automation. Runner H is powered by H’s proprietary compact large language model (LLM) with just 2 billion parameters, emphasizing efficiency over scale.
Access and Features
H has launched a waitlist for Runner H on its site. CEO Charles Kantor confirmed that APIs for pre-built and customizable agents will soon be available, accompanied by H-Studio, a platform for testing and managing these services. Initially, the APIs will be free, with a paid model introduced later.
Funding and Market Strategy
Even compact LLMs are costly to build and operate, especially in a competitive landscape. Kantor highlighted this as part of the “second era of AI,” distinct from companies like OpenAI that defined the first era. H is currently raising a Series A to fuel its ambitious plans.
Since its initial funding, H has reportedly raised an additional $10 million, bringing its total funding to $230 million. Investors include notable individuals like Eric Schmidt, Yuri Milner, and Xavier Niel, along with firms such as Accel and Creandum, and strategic backers like Amazon, Samsung, and UiPath.
Use Cases
H has worked with early adopters in sectors like e-commerce, banking, insurance, and outsourcing to refine Runner H. Customer feedback has driven its development, with initial focus on:
- Robotic Process Automation (RPA): Runner H uses AI to automate repetitive tasks, such as processing forms and moving data across platforms, even when templates change—an area where traditional RPA often falls short.
- Quality Assurance: Aimed at reducing maintenance burdens, Runner H automates tasks like website testing, simulating user actions, and ensuring payment compatibility.
- Business Process Outsourcing (BPO): Runner H streamlines tasks like billing process improvements and data access for agents, enhancing efficiency across operations.
Compact but Competitive
Unlike many foundational AI models with billions of parameters—such as GPT-4 (175 billion)—Runner H takes a streamlined approach with just 2 billion parameters for both its LLM and a computer-vision-based visual language model (VLM). Kantor argues this compact design makes the model more cost-efficient and operationally sustainable, which is critical for retaining business clients.
The company claims its compact model outperforms competitors like Anthropic, Mistral, and Meta in key benchmarks such as WebVoyager. For example, Runner H exceeds Anthropic’s performance in the “Computer Use” category by 29%.
Building for the Future
H sees Runner H as a cornerstone for the “agentic era” of AI, emphasizing adaptability and efficiency in real-world applications. The company’s trajectory signals a focus on practical innovation, customer-centric development, and competitive differentiation.
