Zuckerberg Opens Meta's Most Powerful AI Model to Counter OpenAI Dominance

Meta Platforms (META-US) has announced it will open the weights of its most powerful artificial intelligence (AI) models, directly challenging leading AI labs such as OpenAI and Anthropic that operate under closed models. Meta’s stock rose 2.1% in pre-market trading on Monday.

CEO Mark Zuckerberg announced via an Instagram video on Monday (the 10th) that Meta will release the weights of its latest AI model, Muse Spark 1.2, allowing public download and use. Model weights are the set of parameters and rules that determine how an AI operates and responds.

Meta also unveiled a new open model series called Muse Glimmer, designed to run directly on consumer devices like laptops without relying entirely on expensive cloud data centers.

While OpenAI and Anthropic primarily adopt closed models, Chinese firms such as Alibaba (BABA-US)(09988-HK), DeepSeek, and Moonshot have aggressively released open-weight models in recent years, with some products rivaling top U.S. technologies.

On Monday, Zuckerberg published a 6,500-word article positioning Meta’s move as America’s counterattack against China’s open AI advancements, urging Washington to reduce policy barriers for U.S. companies developing open models.

He argued that foreign AI labs currently hold advantages because U.S. companies face stricter constraints on training data usage. For the U.S. to maintain long-term leadership in open models, policies must reduce these additional frictions.

Zuckerberg opposes restricting access to foreign open models, asserting that America’s true goal should be building the world’s best open models rather than blocking developers from using Chinese technology.

Neil Shah, co-founder of Counterpoint Research, noted that if Western tech giants only build closed 'walled gardens,' developers and enterprises will naturally shift toward Chinese open-weight models. With most U.S. competitors still adopting closed strategies, there is massive demand for non-Chinese open models—offering Meta a prime opportunity to fill this gap.

Muse Glimmer’s on-device design is a key differentiator for Meta. While most AI services rely on cloud data centers, running AI directly on PCs or smartphones reduces cloud computing costs, shortens response times, and could help Meta challenge Google and Microsoft at the device level.

Zuckerberg criticized the concentration of AI power, indirectly targeting OpenAI and Anthropic. He questioned why anyone would rush to create a future where AI eliminates most jobs and diminishes human importance. He also challenged the logic that AI is too dangerous to decentralize, calling extreme centralization fundamentally flawed.

Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have previously warned about AI’s impact on employment, though Altman has recently softened his stance.

Zuckerberg advocates for widespread distribution of AI capabilities rather than concentrating superintelligence. He envisions a future where everyone has a powerful personal AI agent that understands their goals, values, and identity.

Meta’s capital expenditure this year is estimated to reach up to $145 billion, yet its stock has fallen about 10% year-to-date. As investors scrutinize its massive AI spending and competitive edge, Zuckerberg aims to demonstrate progress from the Meta Super AI Lab established last year, proving these investments will eventually yield returns.