Mininglamp Technology Open-Sources Mano-P 1.0 Vision-Based Agent Model

Apr 15, 2026
Mininglamp Technology has open-sourced Mano-P 1.0, a visual AI agent capable of operating desktop and web applications purely through vision, achieving state-of-the-art results across 13 benchmarks and supporting local execution for data privacy.

Mininglamp Technology has officially open-sourced Mano-P 1.0, a GUI-aware agent model that performs complex cross-platform tasks using visual understanding alone, announced in a press release. The model can autonomously control desktop software, web interfaces, and graphical workflows without relying on APIs or HTML parsing.

Mano-P achieved state-of-the-art results across 13 multimodal benchmarks, including ranking first globally in the OSWorld benchmark with a 58.2% task success rate. The 72B version leads in several benchmark systems, while the 4B quantized model runs locally on Apple M4 Pro devices with a 4.3GB memory footprint and speeds of up to 476 tokens per second.

The model supports local deployment for enhanced privacy, running fully offline on Mac devices or via a dedicated USB 4.0 compute stick to ensure data sovereignty. Mano-P is released under the Apache 2.0 license, enabling commercial use and secondary development. Mininglamp is initially open-sourcing the Mano-CUA core skill, with local models and SDKs expected to follow within the month.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Daily AI Brief.

Also, consider following us on social media:

AI Brief AI Brief (X)
Free newsletter

Daily AI Brief

Daily report covering major AI developments and industry news, with both top stories and complete market updates

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.