TLDR: Enhans, a South Korean startup specializing in vertical commerce AI agents, has secured the third position globally in the Online Mind2Web benchmark for web-based AI agents with its ACT-1 model. This places them directly behind industry leaders OpenAI and Anthropic, highlighting their prowess in performing complex web tasks and their leading position in the commerce domain.
Enhans, a burgeoning South Korean startup focused on vertical commerce AI agents, has made a significant global impact by ranking third in the prestigious Online Mind2Web benchmark. Their artificial intelligence model, ACT-1, demonstrated exceptional capabilities in emulating human behavior to perform intricate tasks across the web. This achievement positions Enhans directly behind global tech giants OpenAI and Anthropic, underscoring the advanced nature of their AI technology.
The Online Mind2Web benchmark is a rigorous evaluation that assesses AI agents on their ability to navigate and interact with diverse web environments. The assessment covers 300 real-world tasks spanning 136 actively operating websites, providing a comprehensive measure of an AI agent’s effectiveness in practical scenarios. Enhans notably secured the No. 1 ranking specifically within the commerce domain among all global competitors, showcasing the strong potential of Korean-developed AI agents on the international stage.
Looking ahead, Enhans plans to further evolve its ACT-1 model into a next-generation agent. This evolution will involve integrating imitation learning, a technique that will enable the AI to perform tasks traditionally limited to human users. Additionally, the company intends to expand its enterprise-level web automation solutions. This strategic move aims to differentiate Enhans from other global AI firms and establish a unique competitive edge in the market.
Lee Seung-hyun, CEO of Enhans, expressed the profound significance of this achievement, stating, “Being able to stand alongside global AI leaders like OpenAI and Anthropic is profoundly meaningful. We will continue to strengthen the development of our web AI agent technologies to lead the global market in vertical industries.”
Also Read:
- OpenAI’s ChatGPT Agent Demonstrates Advanced Capabilities by Bypassing CAPTCHA
- South Korea Accelerates Towards AI Leadership by 2030 with Major Investments in Data, Talent, and Infrastructure
ACT-1 is described as a behavior-based AI model specifically tailored for web automation. It is capable of executing sophisticated actions such as clicking buttons and navigating menus. Enhans emphasized that, unlike conventional approaches that often rely on screenshots and coordinate-based controls, ACT-1 utilizes a document object model (DOM)-based method. This allows the model to directly access and manipulate webpage structures, providing a more robust and adaptable approach to web automation.
