Meta released Muse Glimmer, a new open-weight AI model that's far smaller than rival frontier models. It's built for agentic tasks and runs on a Mac or PC with a single graphics card, targeting demand for on-device AI. The launch follows Meta's formation of an expensive superintelligence team last year to push back into the top tier of the AI race.

Alongside the release, Zuckerberg called for lower US barriers on open-source AI to compete with China, arguing foreign labs hold advantages because American labs face more restrictions on training data, and that US policy must cut that friction for American open-source models to lead. He added that blocking access to foreign open-source models isn't an effective solution — the competitive context being that Chinese startups currently lead the open-weight race, with Moonshot's Kimi K3, Alibaba's Qwen3.8-Max, and DeepSeek's V4-Flash rivaling top US systems, while OpenAI, Anthropic, and Google keep their leading models closed.

The article also notes the momentum behind open-weight AI: businesses are wary of ballooning AI bills and recent cybersecurity incidents, and Hugging Face said it used a Chinese open-weight model to defend against a hack because closed models restrict cybersecurity use. Two policy details close it out: the Trump administration reportedly won't put open-weight models through voluntary safety tests, and Zuckerberg said Meta will give independent directors power to approve safety criteria for model releases — while also advocating for distillation, using a powerful AI system to train a smaller one.

Bottom line: this is Meta's two-front play — ship small, cheap, on-device open models to win developers, and lobby Washington so the rules favor that strategy over the closed-model labs.

👉 Have questions? Just reply to this email or contact us at aihelp@axdavid.com - we reply within 24 hours.

​
​Unsubscribe · Preferences​