#China #Beijing #AI Technology #Mininglamp Technology #Mano-P 1.0

Mininglamp Technology Launches Mano-P 1.0: The Future of AI

On April 15, 2026, Mininglamp Technology made headlines by officially releasing Mano-P 1.0, a self-developed model designed to bridge the gap between visual processing and intelligent automation. This groundbreaking GUI-aware agent model distinguishes itself by executing complex cross-platform tasks purely through visual understanding, eliminating traditional barriers of automation relying heavily on API calls and HTML parsing.

Breaking the Bars of AI Limitations

Mano-P 1.0 epitomizes a significant leap in AI capabilities, addressing the long-standing challenge of the 'see but not do' aspect that has plagued many AI systems. By employing a cutting-edge, pure visual comprehension framework, Mano-P allows users to interact with desktop applications, web interfaces, and intricate graphical workflows seamlessly. This revolutionary approach not only enhances user experience but also empowers individuals to design their AI solutions while ensuring data sovereignty—a crucial factor in today's digital landscape.

Performance That Sets New Standards

In the realm of AI benchmarks, Mano-P has redefined excellence. It boasts impressive results across 13 recognized multimodal benchmarks, achieving what can be termed as State-of-the-Art (SOTA) outcomes. The model’s dual-version architecture includes a 72B variant, which has secured the top position in the OSWorld benchmark with a remarkable 58.2% task success rate, significantly outpacing its closest competitor by 13.2 percentage points. A lightweight 4B version also stands out, particularly optimized for local performances on devices such as the Apple M4 Pro, maintaining high efficiency with a mere 4.3GB memory usage while handling up to 476 tokens per second.

Emphasis on Data Privacy

As AI continues to weave itself into crucial business processes, data privacy emerges as a top priority. Mano-P fortifies this aspect by operating entirely offline, devoid of any cloud data uploads. This is made possible by executing the model locally, either on Mac devices or through a dedicated USB 4.0 compute stick. This infrastructure not only protects sensitive data but also allows AI to manage complex processes securely, even in high-security enterprise environments where privacy cannot be compromised.

Promoting a Rich AI Ecosystem

The launch of Mano-P is accompanied by its integration into the AI community, as the software is released under the Apache 2.0 license, allowing for broad and unrestricted usage. Mininglamp Technology aims to provide an effortless entry into the world of AI automation by offering pre-configured mechanisms that require no complex API setups. The first core skill—Mano-CUA—is already being introduced, making it easy for developers to incorporate it into platforms like OpenClaw or Claude Code. Upcoming updates promise advancements in training methodologies and token pruning technologies, propelling the growth of a personalized AI ecosystem on a global scale.

Conclusion

With the unveiling of Mano-P 1.0, Mininglamp Technology is set to revolutionize the approach toward AI functionalities, emphasizing personalization, performance, and unparalleled data protection. This initiative not only showcases the company’s commitment to innovation but signals a transformative shift in how AI-powered tools are developed and utilized in various sectors.