(L-R) Nvidia Corporation President and CEO Jensen Huang, Google CEO Sundar Pichai, and Meta CEO Mark Zuckerberg attend the U.S. Senate Bipartisan Artificial Intelligence Insights Forum at the U.S. Capitol in Washington, DC, on September 13, 2023.
Andrew Caballero-Reynolds | AFP | Getty Images
Last month, US tech giants came together to urge policymakers not to impose “premature restrictions” on promiscuous AI models, even those made in China. Two of these companies are now making a joint effort to compete by introducing their own public offerings.
Meta and Nvidia The companies this week released artificial intelligence models that developers can download for free via the open source ecosystem. OpenAI and human.
Open source AI has been the subject of controversy from Silicon Valley to Washington, D.C., with critics raising concerns about the potential national security risks of the Chinese model and an AI training practice called distillation that is considered a form of intellectual property theft. Meanwhile, most major players in the industry argue that restricting the use of the model would be to their own detriment and place too much power in the hands of too few companies.
“The age of AI can be an era of prosperity,” the consortium of tech companies said in a July 24 open letter. “With the right choices, open-weight AI can expand opportunity, strengthen competition, extend America’s technological leadership, reduce risk, and ensure the benefits of this extraordinary technology are shared broadly across the economy.”
They call distillation “a widely used technique for model improvement, evaluation, and validation.”
Meta released Muse Glimmer on Monday as part of its strategy to release its most powerful AI models to the open source community. CEO Mark Zuckerberg said the company will publish the weights of its latest AI model, Muse Spark 1.2. Weights refer to the calculations and rules that determine the behavior and behavior of the AI.
The next day, Nvidia debuted Nemotron 3.5 Lightning. This model is derived from the company’s Nemotron 3 family of models released in December. The chipmaker said its models are “true open source” because it publishes the relevant “training datasets, techniques, and model weights” for developers to inspect.
Both companies still need to prove that their products have an audience in a market that features popular models from Chinese AI labs like Moonshot AI and DeepSeek. alibaba’s Kwen.
box CEO Aaron Levy, one of the signatories of last month’s letter, is optimistic. He said Zuckerberg’s Muse Spark 1.2 plan is a “huge deal” because it is a powerful model comparable to Anthropic’s and OpenAI’s top foundation models. This week’s models from Meta and Nvidia are small and intended to run on laptops for tasks such as powering up digital agents on devices.
“The flag is firmly in place that there will be a near-frontier open source model in America,” Levy said.
Meta has tried this route with Llamas in the past. This was Zuckerberg’s first foray into the basic AI market, but developers were not impressed with the release of Llama 4 in April 2025. Meta then spent billions of dollars overhauling its AI division and installed Scale AI CEO Alexandr Wang to lead the division.
Recently, Wang’s group has been rolling out its own model under the Muse brand, aiming to develop new revenue streams.
“True potential”
Levy said companies hesitant to use China’s open-weight AI models will be more likely to experiment with future variants of Meta.
“As an example, it would probably be impossible to implement a non-domestic open source model in a major government agency, and it would probably be impossible to use it in a very large bank,” Levy said. “When you think about what use cases Muse can be used in, it really opens up a huge amount of possibilities.”
Still, Meta faces headwinds, especially as it pursues yet another open source strategy. Umesh Sachdev, CEO of business AI startup Uniphore, said Meta bridged the gap with third-party developers as they transitioned from open weight to proprietary AI models.
“I think it’s going to take a 3,500-plus word article from Zack to convince developers,” Zuckerberg said of the manifesto Zuckerberg released this week. “Uniphore developers feel almost betrayed.”
However, Sachdev said he is rooting for domestic companies to succeed because “more competition lowers token costs and fosters innovation, which is always good for consumers.”
This is a sentiment echoed by Forrester analyst Charlie Dye. He called Meta’s latest move “strategically important as it brings America’s leading frontier AI vendors back into the open ecosystem.”
“Developers and enterprises will welcome Meta’s transition to open weight as it increases transparency, customization, deployment flexibility, and data sovereignty,” said Dai. Going forward, the company “must prove that it can not only release competitive models, but also foster a durable ecosystem,” he said.
Spotlight: Meta’s new AI model is a “positive development for the ecosystem.”
