A Chinese startup has unveiled Manus, touting it as the world’s first fully autonomous AI agent capable of independently handling real-world tasks. This groundbreaking innovation has achieved new state-of-the-art (SOTA) performance on agentic benchmarks, setting a new standard for AI capabilities.
In a the demo, Manus showcased its ability to perform complex tasks such as resume screening and property research. What sets it apart is its access to an independent computer instance, enabling it to execute tasks seamlessly. The AI agent also demonstrated advanced skills, including web browsing, coding, and creating visuals. Impressively, it can even manage tasks on platforms like Upwork and Fiverr, marking a significant leap in AI autonomy.
Manus has outperformed leading general-purpose AI assistants like ChatGPT and Gemini on the GAIA benchmark, a comprehensive evaluation of AI performance. This achievement highlights its potential to redefine how AI handles complex workflows.
Currently, Manus operates on an invite-only basis, but the team has committed to open-sourcing the models behind the agent later this year. This move is expected to accelerate innovation and adoption across industries.
We are witnessing a pivotal moment in AI development, where relatively unknown labs are introducing state-of-the-art tools. While early AI agents required human supervision for simple tasks, Manus represents the next step: fully autonomous agents capable of managing intricate workflows independently.