Manus (AI agent)
 | This article needs additional citations for verification. (March 2025) |
| Developer(s) | Monica (also known as Butterfly Effect AI) |
|---|---|
| Initial release | March 6, 2025 |
| Type | |
| License | Proprietary |
| Website | manus |
Manus (hand in Latin) is an autonomous artificial intelligence agent developed by Chinese startup company Monica.[1] The agent is designed to independently carry out complex online tasks without direct/continuous human guidance.
History
[edit]Manus was founded to create artificial intelligence agents capable of operating independently, based on large language models (LLM). The official launch of Manus on March 6, 2025, drew international attention. Experts and media described Manus as a major advance because it could autonomously handle complex tasks, including writing and deploying code, without direct human intervention.[2]
Performance
[edit]Manus is claimed as a fully autonomous AI agent, designed to handle tasks like website creation, stock analysis, travel planning, and schedule management.[3] It has demonstrated performance on the GAIA benchmark, a test of real-world problem-solving skills, with reports indicating a score around 86.5%, potentially exceeding models like H2O.ai’s h2oGPTe Agent (65%) and OpenAI’s DeepResearch, though exact figures remain unverified due to limited public data.[4][5][6][7] In practical use, Manus can compile lists, search properties, and nominate candidates, sometimes performing better than competitors like ChatGPT DeepResearch in specific tasks, but it requires user intervention for paywalls and captchas and can take longer, with task times ranging from 30 minutes to over an hour. Its transparency, displaying its process through a side panel, makes it accessible for users regardless of coding experience, and it operates asynchronously in the cloud, allowing tasks to proceed without constant supervision.[8][9] But it encounters challenges with system stability, including crashes and server overload under high demand, leading to task failures and a higher failure rate than some alternatives. Access is limited, with fewer than 1% of waitlist users receiving invite codes, and its $2 per-task cost is positioned competitively but affected by reliability considerations. While recognized for its capabilities and seen as innovative by some, Manus has received scrutiny for potentially being a wrapper of existing models and for its marketing approach, as server issues and limited availability raise questions.[10]
See also
[edit]References
[edit]- ^ Fried, Ina (2025-03-10). "New Chinese AI agent Manus draws DeepSeek comparison". Axios. Retrieved 2025-03-10.
- ^ Franzen, Carl (2025-03-10). "What you need to know about Manus, the new AI agentic system from China hailed as a second 'DeepSeek moment'". VentureBeat. Retrieved 2025-03-13.
- ^ Chen, Caiwei (March 11, 2025). "Everyone in AI is talking about Manus. We put it to the test". MIT Technology Review. Retrieved 2025-03-28.
- ^ "H2O.ai Tops GAIA Leaderboard: A New Era of AI Agents". h2o.ai. Retrieved 2025-03-28.
- ^ Mialon, Grégoire; Fourrier, Clémentine; Swift, Craig; Wolf, Thomas; LeCun, Yann; Scialom, Thomas (2023-11-21), GAIA: a benchmark for General AI Assistants, arXiv, doi:10.48550/arXiv.2311.12983, arXiv:2311.12983, retrieved 2025-03-28
- ^ "GAIA Leaderboard - a Hugging Face Space by gaia-benchmark". huggingface.co. Retrieved 2025-03-28.
- ^ "Manus AI: The Best Autonomous AI Agent Redefining Automation and Productivity". huggingface.co. 2025-03-06. Retrieved 2025-03-28.
- ^ Smith, Craig S. "China's Autonomous Agent, Manus, Changes Everything". Forbes. Retrieved 2025-03-28.
- ^ "Introducing Manus: The general AI agent — WorkOS". workos.com. Retrieved 2025-03-28.
- ^ Chen, Wency (2025-03-07). "Chinese AI agent Manus faces doubts after rapid rise to fame". South China Morning Post. Retrieved 2025-03-28.
