Meta is doing something rare in the AI world: giving away the keys.
On Monday, Aug 10 2026, CEO Mark Zuckerberg said the company will open source its most capable AI models. That includes Muse Spark 1.2, the newest version of its top model.
Meta is releasing the model weights for Muse Spark 1.2, allowing anyone to download and access them. Weights are the internal settings that control how an AI model behaves, like the recipe inside a black box.
Meta is also launching a new open-source model family called Muse Glimmer, built for laptops. The goal is to put useful AI directly on a device instead of routing everything through a distant data center.
This move builds on Meta's earlier work with the Llama series, which made open-weight models a cornerstone of its AI strategy. The creation of Meta Superintelligence Labs last year signaled a deeper commitment to frontier research, and today's announcement marks the first time the company has opened the weights of its most advanced model.
OpenAI and Anthropic generally keep their models closed. Meta is going the opposite way, and Zuckerberg is making a political case for it.
In a 6,500-word essay published Monday, he argued that American labs face extra restrictions on training data. He argued that overseas competitors benefit from fewer constraints, while U.S. companies face additional rules around training data.
"US policy must reduce this additional friction if we want American open source models to lead over time," he wrote.
He also focused on competition from China. Alibaba, DeepSeek and Moonshot have already released open-weight models that compete in certain areas, and Zuckerberg wants American open-source models to be the best globally.
Get the free Always Be Buying eBook and learn the simple system for building wealth on any income
Zuckerberg also warned against letting a small number of companies control AI. That comment was widely seen as a jab at OpenAI and Anthropic.
"Rather than centralizing superintelligence, we should distribute it widely and give every person the ability to direct it," he said.
He even pushed back on the doom talk around AI, which has come from the heads of Anthropic and OpenAI, Dario Amodei and Sam Altman.
Altman has recently softened his tone, and Zuckerberg says he does not understand why anyone who thinks AI will eliminate most jobs would rush to build that future.
Meta's announcement is about more than AI philosophy. The company is spending heavily to back it up, with capital spending estimated at as much as $145 billion this year.
The spending is supposed to show that Meta Superintelligence Labs, created last year, is making real progress. Investors are not fully convinced. A big part of the story is whether that spending pays off.
Analysts see a practical reason for the strategy. If Meta can put smaller AI models directly on laptops and phones, those models can process tasks on the device.
That cuts cloud computing costs and speeds up response times. Counterpoint Research's Neil Shah said it could let Meta beat Google, Microsoft and others on the device itself.
Shah also said that if Western tech giants only build closed platforms, or walled gardens, developers will naturally pivot to Chinese open-weight models. He said Meta can step in because developers outside China are hungry for open-weight models that are not Chinese.
The bigger picture is a fight over who controls AI. Zuckerberg says that distributing superintelligence widely would empower individuals and give everyone a personal AI agent.
That sounds convenient, and it is also a business strategy. The more people and companies use Meta's models, the more products Meta can sell around them.
For your portfolio, the question is whether Meta's giant bet pays off. If open-source AI takes off, Meta could profit by putting its models on all kinds of devices and building services around them.
If that spending does not turn into products people want, $145 billion is a lot of money to watch evaporate. That risk helps explain the 10% stock drop this year.
For everyday life, the upside is easy to picture. AI that runs on your laptop, without waiting on a server, can feel faster and cheaper.
You may end up with a personal assistant that understands what you care about. Who builds that assistant, and how open it really is, may be the biggest question behind this week's news.
Download the free Always Be Buying eBook and start putting your money to work today