Meta opens up its strongest AI models, taking a jab at OpenAI and Anthropic

Meta has announced it will make its most advanced AI models open source, a move aimed at competing with major labs like OpenAI and Anthropic. In a video posted Monday, CEO Mark Zuckerberg said the company would release the weights for its latest model, Muse Spark 1.2, allowing the public to download and use it. Weights are the internal calculations that guide how an AI behaves.

The company also plans to launch a new family of open-source models called Muse Glimmer, which are built to run on a laptop. This on-device approach differs from rivals who rely on cloud-based processing, potentially lowering costs and speeding up AI on consumer electronics.

Meta's shares rose 2.1% in premarket trading Monday, though they remain down about 10% this year. Investors have been watching Meta's heavy spending, with capital expenditure projected at up to $145 billion for the year. Zuckerberg is trying to reassure them that the Meta Superintelligence Labs, formed last year, is making progress.

In a lengthy essay published Monday, Zuckerberg positioned the move as a challenge to Chinese open-source technology and called on Washington to reduce restrictions on training data for American labs. He argued that limiting access to foreign open models is not effective and that U.S. open-source models should be the best globally.

Analysts note that while U.S. companies like OpenAI and Anthropic focus on closed models, Chinese firms have aggressively released open-weight versions. One analyst said Meta can fill a void for non-Chinese open models, which are in high demand.

Zuckerberg also warned against concentrating AI power in a few companies, appearing to take a swipe at OpenAI and Anthropic. He said that distributing superintelligence widely could empower individuals. Both OpenAI's Sam Altman and Anthropic's Dario Amodei have previously warned about AI's impact on jobs, though Altman has recently softened his stance.

Source: CNBC