Meta and Nvidia have released AI models with open weights to compete with China
Meta and Nvidia unveiled open-weight AI models this week, available to developers for free download through the open-source ecosystem. U.S. tech companies are trying to step up competition with Chinese AI labs, including DeepSeek, Moonshot AI, and Alibaba with its Qwen models, according to CNBC.
On Monday, Meta released the Muse Glimmer model as part of its strategy to provide open access to its most powerful models. The company’s CEO, Mark Zuckerberg, also announced plans to release the weights for the new Muse Spark 1.2 model. Weights are the computational parameters and rules that determine the operation and behavior of an AI model.
The following day, Nvidia unveiled Nemotron 3.5 Lightning—a model based on the Nemotron 3 family, which was released in December. The chipmaker has called its models truly open-source, as it publishes training datasets, methodologies, and model weights for developers to verify.
For more breaking news, follow the UA.News Telegram channel.
Box CEO Aaron Levy, who signed an open letter from tech companies, considers Meta’s plans for Muse Spark 1.2 to be very important. According to him, this model can compete with the leading base models from Anthropic and OpenAI. At the same time, the models presented this week by Meta and Nvidia are smaller and designed to run on laptops, specifically for running digital agents directly on the device.
On July 24, more than 20 U.S. tech companies issued an open letter urging policymakers not to impose premature restrictions on models with open weights, particularly those developed in China. The authors of the letter stated that such models could expand opportunities, strengthen competition and U.S. technological leadership, and ensure a broader distribution of the benefits of AI across the economy.
Forrester analyst Charlie Dai called Meta’s move strategically important, as it brings a major U.S. provider of cutting-edge AI solutions back into the open ecosystem. In his view, developers and companies may welcome Meta’s return to open weights due to greater transparency, customization options, deployment flexibility, and data sovereignty.
