Genspark’s ‘Super Agent’ fundamentally changes the framework of existing AI assistants. Moving beyond merely executing given commands, it understands the user’s purpose, plans tasks, and independently executes complex processes as a general-purpose agent. It has attracted industry attention for applying the ‘MoA (Mixture-of-Agents)’ system, a collaborative structure among AI, breaking away from a single-model focus.
Genspark Super Agent goes beyond simple Q&A by autonomously handling complex tasks based on user demands. For instance, if a travel plan is requested, it collects relevant information, compares transportation and lodging options, and even makes reservation calls to complete the schedule. It summarizes a 5-hour video into 10 slides and automatically generates a visualized analysis report after researching global industry trends.
The content creation function has also expanded. The system automatically processes tasks ranging from creating 1-minute news-based animations to social media videos using step-by-step recipes and providing shopping recommendations within a specific budget range. The breadth of its applications is extensive, thanks to its ability to understand user needs and deliver results that reflect them thoroughly.
The Super Agent’s capabilities are diverse. In the released demo video, scenes were shown where the AI completed restaurant reservations on behalf of the user, compressed long video content into 10 summary slides, and created a global market report alongside data visualization. Additionally, it automatically produced 1-minute episodes in the style of ‘South Park’ based on news content, Instagram Reels videos, marketing posters, and even 3D visualizations of mathematical formulas. This makes it a versatile AI for content creators, marketers, researchers, and individual users alike.
The core of Super Agent lies in Genspark’s world-first introduction of the MoA (Mixture-of-Agents) system. It assigns multiple AI models like GPT-4o, Claude, and Gemini to different tasks, utilizing each model’s strengths, reducing error possibilities through cross-verification, and providing more accurate and reliable outcomes.
The MoA system is akin to experts from various fields collaborating for a single project within a company. For example, when asked to analyze the music industry, one model acts as a researcher collecting data, another as a designer composing visual materials, and the final model as a content planner summarizing the content. These AI roles work concurrently while the entire process is coordinated by a central system to deliver a well-integrated result.
In the GAIA benchmark, recognized as a global AI performance evaluation standard, Genspark Super Agent surpassed existing models at all levels. It recorded the highest accuracy, scoring 87.8% at Level 1, 72.7% at Level 2, and 58.8% at Level 3, surpassing renowned research institutions like OpenAI and manus.ai.
Genspark has launched the Super Agent through an iOS app and is preparing an Android version. Along with this announcement, they secured $100 million in Series A funding, officially starting the construction of an AI-based super app ecosystem.
Super Agent not only signifies technological progress but also heralds the domestication of AI. It enables delegating repetitive tasks and focusing more on creative work, designed for easy access to advanced AI technology by anyone.
Now, AI is evolving from a mere helper to a ‘producer, thinker, and actor’ replacing human roles.
