Urgent.News

What's breaking now, across thousands of outlets.

AI

Meta brandishes open-source AI models again, as Zuckerberg media blitz emphasizes battle against Chinese rivals

It comes as Meta tries to regain ground on OpenAI and Anthropic, and as Chinese open-source labs close in on the American frontier.

Meta brandishes open-source AI models again, as Zuckerberg media blitz emphasizes battle against Chinese rivals

Meta introduced its latest generation of AI models, Magma Glimmer, which are open-source and capable of running on a laptop, on Monday. Accompanying the announcement was a lengthy essay by Meta CEO Mark Zuckerberg, who argued for the advantage of open-source AI over controlled access models primarily owned by a few companies. Zuckerberg highlighted that powerful AI should not be dominated by a select few, referring to competitors such as OpenAI and Anthropic as though they were directly at odds with Meta.

He emphasized Meta's commitment to open-source as a crucial factor in challenging Chinese open-source models, specifically DeepSeek and Moonshot, which have been approaching the American market. Zuckerberg also called for increased support from Washington to help American open-source efforts. As part of this new initiative, Meta intends to release the weights of Muse Spark 1.2, its most advanced model, allowing anyone to download, examine, and alter it instead of only being able to access it through Meta's products or API.

Zuckerberg's argument is part of a coordinated effort by Meta to promote American open-source models as globally superior, a goal that requires dismantling barriers that make it difficult for American models to compete. The release of Magma Glimmer comes at a time when the American AI market is dominated by closed-source companies like OpenAI, Anthropic, and Google, which keep their most advanced "frontier" models inaccessible to the public.

Meta originally aimed to stand out by embracing open-source, launching the open-weight Llama family of models in February 2023, which allowed developers outside of Meta to download and modify high-performing models. However, Meta's commitment to open-weight has become less consistent, with the company now mixing open and proprietary releases.

Over the past year, Meta has invested heavily in Superintelligence Labs, hiring former Scale AI CEO Alexander Wang to help position the Magma family as a foundation for an advanced Meta AI assistant across its various platforms. This shift towards open-source may also be Meta's attempt to regain its competitive edge following the release of less impressive models compared to those from OpenAI and Anthropic.

By emphasizing open-source, Meta intends to differentiate itself from rivals that keep their most capable systems behind closed doors and garner political support from users and the government.

Written by urgent.news from Fortune's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at fortune.com →

More in AI

More from Monday 10 August →