SAN FRANCISCO — Meta and Nvidia released new artificial intelligence models this week, making them available for free download to developers and pushing them directly into the open-source ecosystem against proprietary rivals.
The releases follow a lobbying push by American tech firms. On July 24, a consortium that included Meta and Nvidia sent an open letter urging policymakers to avoid premature restrictions on open-weight AI models. The letter argued that open-weight AI expands opportunity, strengthens competition and extends U.S. technological leadership, and said such models mitigate risk and broaden access to technology benefits.
The open-weight market faces strong competition from Chinese labs. Moonshot AI, DeepSeek and Alibaba's Qwen all have popular models in active use.
Meta CEO Mark Zuckerberg unveiled Muse Glimmer as part of a strategy to release powerful AI models to the open-source community. He also committed to opening the weights for Muse Spark 1.2, Meta's latest AI model. Weights are the calculations and rules that define how an AI system functions.
A day later, Nvidia debuted Nemotron 3.5 Lightning, an extension of its Nemotron 3 model family released in December. Nvidia said its models are fully open source because the company publishes training datasets, techniques and model weights for developer inspection.
Both models are compact, designed to run on laptops rather than in large-scale data centers. They are built for tasks such as powering on-device digital agents.
Box CEO Aaron Levie, a signatory of the July 24 letter, said Zuckerberg's plan for Muse Spark 1.2 is a "very big deal" because the model rivals top foundation models from Anthropic and OpenAI. He said the releases plant a "very firm flag in the ground that America will have near-frontier open-source models."
Meta has pursued open-source AI before with its Llama series. The release of Llama 4 in April 2025, however, left developers unimpressed, prompting Meta to spend billions of dollars overhauling its AI infrastructure.
Open-source AI remains contentious in Silicon Valley and Washington. Critics raise concerns about national security risks from Chinese models and intellectual property issues tied to AI training practices such as distillation. The tech consortium described distillation as a "widely used technique for model improvement, evaluation, and validation" — a position that clashes with critics who view it as intellectual property theft.
Meta and Nvidia must now prove there is a significant audience for their open-source offerings, competing against established models from Chinese rivals.
