Meta has unveiled its latest bid to reshape the AI landscape by focusing on open-weight large language models. The company’s CEO, Mark Zuckerberg, dived into a lengthy essay to differentiate Meta from rivals like OpenAI and Anthropic, both of which have staunchly defended their closed-source strategies.

The announcement includes the launch of an open-source model named Muse Glimmer and plans to release the weights for its more powerful Muse Spark 1.2 in the coming weeks, marking a significant shift towards transparency and collaboration within the AI community.

Muse Glimmer, with its 30 billion parameters and ability to run locally on users’ machines, is poised to compete not just with other open-weight models but also against proprietary systems from Chinese labs. Despite generally lower capabilities compared to leading models like Anthropic’s or OpenAI's, it positions itself as a more affordable alternative.

Meta’s move towards openness comes at a time when smaller, more efficient AI models are becoming increasingly popular. As such models gain traction, Meta aims to reduce reliance on the big labs by empowering users with local inference capabilities, potentially cutting costs and enhancing privacy.