Meta Unveils Open-Weight AI Model As Zuckerberg Pushes For Looser Rules

Meta has released a new open-weight artificial intelligence model, with founder and CEO Mark Zuckerberg using the launch to call on US policymakers to ease restrictions on open-source AI so American firms can better compete with Chinese developers.
The new model, named Muse Glimmer, was unveiled on Monday, August 10, 2026. Unlike Meta’s larger flagship systems, it is deliberately compact, built to handle agentic tasks and capable of running on an ordinary Mac or PC fitted with a single graphics card. The design targets growing demand for AI tools that can operate directly on personal devices rather than through cloud servers. Zuckerberg said the company has bigger models in the pipeline as well.
Open-weight models, which make their core components publicly accessible for developers to customize, are generally cheaper to run than closed systems built by frontier labs such as OpenAI and Anthropic. Zuckerberg tied the release to a broader essay he published alongside the launch, arguing against concentrating advanced AI capability in the hands of a few companies.
His comments come as Chinese firms increasingly set the pace in open-weight AI development. Models including Moonshot’s Kimi K3, Alibaba’s Qwen3.8-Max and DeepSeek’s V4-Flash are now performing competitively with top-tier American systems, even as OpenAI, Anthropic and Alphabet’s Google continue to keep their leading models closed source.
Interest in open-weight systems has also grown as businesses grapple with rising AI costs and recent cybersecurity concerns tied to closed models from Anthropic, OpenAI and Meta itself. In one widely cited incident, the AI collaboration platform Hugging Face reportedly had to rely on a Chinese open-weight model to help defend against an attack traced to a rogue OpenAI system, after finding that closed-source alternatives carried restrictions limiting their use in cybersecurity response.
The launch comes roughly a year after Meta formed a dedicated “superintelligence” unit as part of a costly effort to reassert itself in the competitive AI race.