Meta announced Monday it will open source its most advanced AI model yet. The move puts the company in a different lane than rivals OpenAI and Anthropic, which have kept their top models closed to the public.
CEO Mark Zuckerberg shared the news in an Instagram video. He said Meta will release the weights for its latest model, called Muse Spark 1.2. Weights are the internal settings that control how an AI system behaves, and releasing them lets outside developers download and build on the technology for free.
Zuckerberg also revealed a second product launch. Meta is introducing a new family of models called Muse Glimmer. These are designed to run directly on laptops instead of relying on cloud data centers.
Why Meta Is Going Open Source
Most AI labs in the United States keep their strongest models private. Companies in China, including Alibaba, DeepSeek, and Moonshot, have taken the opposite path and released open models that compete with American technology in some areas.
Zuckerberg wrote a long essay on Monday laying out his reasoning. He argued that U.S. labs face extra rules on training data that put them at a disadvantage against foreign competitors.
He said restricting access to foreign open source tools is not the right fix. Instead, he wants U.S. policy changes that make it easier for American open models to lead globally.
Analyst Neil Shah of Counterpoint Research said the strategy fills a gap in the market. He noted that most U.S. competitors keep their models closed, while demand for open, non-Chinese models remains high.
Shah also pointed to the laptop-focused Muse Glimmer models. Running AI directly on a device instead of the cloud can cut costs and improve speed for everyday users.
Investor Reaction and Zuckerberg's AI Vision
Meta shares climbed 2.1% in premarket trading following the announcement. The stock is still down roughly 10% so far this year as investors watch the company's heavy AI spending.
Meta expects capital spending to reach up to $145 billion this year. Zuckerberg is trying to show investors that Meta Superintelligence Labs, formed last year, is producing results worth that price tag.
In his essay, Zuckerberg also took aim at how other AI leaders talk about the technology's risks. He said he does not understand why people who believe AI will eliminate jobs would still rush to build it.
He argued against concentrating AI power in a small number of companies. Both Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have warned publicly about AI's effect on jobs, though Altman has softened his tone recently.
Zuckerberg described his own vision as the opposite of concentration. He said AI should be spread widely so every person can direct it toward their own goals.
He closed his essay by predicting a future where everyone has a personal AI agent. That agent, he said, would understand a person's goals and help them pursue their interests
