Meta Platforms CEO Mark Zuckerberg has advocated for reduced U.S. barriers on open-source AI in order to enhance competitiveness against Chinese counterparts. The tech giant unveiled a new, smaller, and customizable model named Muse Glimmer, intended for tasks on Mac or PC with a single graphics card. Zuckerberg hinted at the imminent launch of larger models, emphasizing the importance of democratizing AI rather than centralizing power.
The push for open-weight AI has gained momentum as companies seek cost-effective alternatives to proprietary models amid concerns over cybersecurity incidents involving models from Anthropic, OpenAI, and Meta. Open-weight models, which allow users to customize core components, stand in contrast to closed models that offer limited flexibility and higher costs.
Recent high-profile Chinese open-weight models like Kimi K3 have spotlighted the benefits of open-source AI. Notably, Meta, alongside industry players such as Nvidia, Microsoft, and Palantir, signed an open letter endorsing open-weight AI. Chinese startups have shown prowess in open-weight models, with offerings like Kimi K3, Qwen3.8-Max, and V4-Flash rivaling leading U.S. systems.
Meta’s stock, which has experienced a 10% decline this year, saw a 3% increase in premarket trading following the Muse Glimmer model’s release. The company plans to unveil Muse Spark 1.2, its most advanced model developed by a dedicated superintelligence team. Zuckerberg highlighted the need for the U.S. to reevaluate policies around open-weight models, data use, and distillation to foster innovation and leadership in the AI domain.
To address concerns over data centre expansion, Meta announced a $1 billion fund to support communities affected by its infrastructure projects. Zuckerberg emphasized the importance of infrastructure development to bolster AI capabilities, acknowledging the challenges faced by U.S. firms compared to their Chinese counterparts. Meta aims to implement governance structures to ensure the safe release of models and drive advancements in open-weight AI.