OpenAGI launched something interesting - Lux, a foundation AI agent that actually operates a computer better than current solutions from OpenAI, Google, and Anthropic.
Benchmark numbers (Online-Mind2Web, 300+ real tasks):
<ul>
<li>Lux - 83.6%</li>
<li>Google Gemini CUA - 69%</li>
<li>OpenAI Operator - 61.3%</li>
<li>Claude Computer Use - 56.3%</li>
</ul>
It also performs actions in ~1 second per step (Operator needs ~3s) and costs 10× less.
But the real unlock: Lux works not just in the browser - it can navigate desktop apps like Excel, Slack, Adobe, IDEs… even Dota 2.
The model was trained not on text, but on screenshots + action sequences, so it generates actions instead of paragraphs.
Available via API &amp; SDK. And OpenAGI is already working with Intel on local optimization for laptops.
<a href="https://www.agiopen.org/" target="_blank" rel="noopener nofollow">https://www.agiopen.org/</a>

Valdemar

OpenAGI released Lux, a foundation AI agent that controls computers through screenshots and action sequences rather than text. It outperforms competing solutions from OpenAI, Google, and Anthropic on real-world tasks (83.6% vs 69% for Gemini CUA), operates faster (~1 second per step), and costs 10× less. Unlike browser-only alternatives, Lux works across desktop applications including Excel, Slack, Adobe products, and IDEs. The model is available via API and SDK, with Intel collaboration underway for local laptop optimization.