Meta Releases Open-Weight Muse Glimmer AI Model Amid Competition From China

Meta has unveiled the Muse Glimmer open-weight AI model, which users can run locally, and plans to release the weights for its frontier model, Muse Spark 1.2. This comes amid competition from Chinese open-weight AI models, with the company’s CEO, Mark Zuckerberg, reiterating the need for these models as the U.S. seeks to lead the AI race.

Meta CEO Announces Launch Of Muse Glimmer

In an X post, Mark Zuckerberg announced that they are opening the weights for Muse Glimmer, which he described as a “great 30B parameter dense model that can run locally.” He also mentioned that they plan to release the weights for Muse Spark 1.2, their latest and most powerful frontier model.

“Meta is a strong supporter of open source and I’m proud of these releases,” Zuckerberg said. As CoinGape earlier reported, Meta was among the tech firms that wrote an open letter warning the Trump administration against restrictions on open-weight models amid increased competition from China.

Chinese open-weight AI models from DeepSeek, Moonshot, and Alibaba have continued to gain hype given how they match up to models from U.S. firms such as OpenAI and Anthropic while serving as a cheaper alternative. The hype around Moonshot’s Kimi K3 last month had led to reports that the U.S. was considering restrictions on Chinese models.

However, Meta and other tech firms such as Nvidia made a case for open-weight models in their open letter while noting that these models can expand opportunity and strengthen competition. Open-weight models are models that users can customize and deploy on their own infrastructure.

Important For The U.S. To Lead Open-Source Ecosystem

In an essay dubbed “The Future is for Everyone,” Zuckerberg stressed the importance of the U.S. and its allies leading the open-source AI ecosystem, which he noted will make up a “large percent of global AI use.”

The Meta CEO asserted that Foreign Labs currently hold “several advantages” in this regard since they do not have to comply with many additional restrictions on training data. “US policy must reduce this additional friction if we want American open source models to lead over time,” he advised.

Echoing the open letter which his company signed, he opined that restricting access to foreign open source models isn’t an effective solution. Instead, he believes the focus should be on making American open source models the best globally, which would require removing the hurdles that make it harder to compete.

Zuckerberg also warned against policies that could slow the release of American frontier models even by a month. In this regard, he called for a collaboration between frontier labs and the government rather than a rigid process and review timeline that is followed in all cases.

To implement AI in your crypto trading strategy, check out Best Crypto AI Trading Bots in 2026: Which Bots Actually Use AI?