Meta Releases Open-Source AI Model to Challenge Rivals
Meta on Monday released Muse Glimmer, an open-weight AI model built for agentic tasks and small enough to run on a Mac or PC with a single graphics card.
- Meta trained Glimmer via distillation from Muse Spark, its most powerful model.
- Meta plans to open-source Muse Spark 1.2, its flagship model, imminently.
- Meta shares were up 2.1% premarket Monday; they are down roughly 10% this year.
- Meta's capital expenditure is forecast at up to $145 billion this year.
Why it matters: Open-weight models can be downloaded, customized, and run locally — making them cheaper than closed models from OpenAI or Anthropic. Demand is rising as businesses face higher AI costs and recent cybersecurity incidents.
- Chinese open-weight models — Moonshot's Kimi K3, Alibaba's Qwen3.8-Max, DeepSeek's V4-Flash — now rival top U.S. lab outputs.
- Hugging Face said it used a Chinese open-weight model to repel a cyberattack because closed-source models restrict cybersecurity use.
How 10 sources split on this story
Where they split: The core dispute is whether loosening U.S. open-source AI policy closes a competitive gap with China or creates national-security risks.
Center coverage, 7 sources: The center focuses on the competitive and policy mechanics — U.S. versus China in open-weight AI, investor scrutiny of Meta's spending, and the specifics of Zuckerberg's regulatory asks.
CNBC50mMeta to open source its most powerful AI model as it takes swipe at OpenAI, AnthropicCNCTV News2hMeta launches new AI model as Zuckerberg champions open-weight push
Reuters26mMeta lança novo modelo de IA e Zuckerberg defende avanço dos modelos de pesos abertos
Bloomberg3hMeta Releases Muse Glimmer AI Model People Can Run on Their Laptop
Engadget44mMeta's 'Open Source' Muse Glimmer Model Can Run On A Single Computer
Financial Times3hMark Zuckerberg attacks ‘closed’ AI rivals as Meta returns to open models
USA Today3hMeta launches new Muse Glimmer AI modelLeft coverage, 2 sources: The left frames the story around Meta's broader vision for open superintelligence and Zuckerberg's 6,500-word philosophical case for AI development.
Right coverage, 1 sources: The right emphasizes the national-security dimension, framing the story as a race against China and noting Washington's existing and proposed restrictions on Chinese AI models.
What’s next: The Trump administration said it will not subject open-weight models to voluntary safety tests, per two sources familiar with the discussions.
- Washington is weighing broader restrictions on Chinese AI models; DeepSeek is already barred from U.S. intelligence systems.
- Will Meta's policy push prompt concrete regulatory changes, or remain a public argument?
- How will OpenAI and Anthropic respond to Zuckerberg's direct challenge to closed-model development?
- Can Meta's open-weight models match Chinese competitors on performance, not just accessibility?