Updated
Updated · CNBC · Aug 12
Meta, Nvidia Release 2 Open-Weight AI Models as U.S. Labs Push Back on China
Updated
Updated · CNBC · Aug 12

Meta, Nvidia Release 2 Open-Weight AI Models as U.S. Labs Push Back on China

3 articles · Updated · CNBC · Aug 12

Summary

  • Meta and Nvidia this week made new AI models free to download through the open-source ecosystem, with Meta unveiling Muse Glimmer and Nvidia launching Nemotron 3.5 Lightning.
  • The releases advance a U.S. push for open-weight AI after a July 24 industry letter warned against restricting such models, arguing openness can strengthen competition and American technological leadership.
  • Meta's next step is bigger: Mark Zuckerberg said the company will open the weights for Muse Spark 1.2, which Box CEO Aaron Levie said could rival top Anthropic and OpenAI foundation models.
  • Both companies are targeting a market where Chinese open-weight models from DeepSeek, Moonshot AI and Alibaba's Qwen are already popular, especially as some enterprises and government users prefer domestic alternatives.
  • Meta still faces skepticism after disappointing developers with Llama 4 and then shifting toward proprietary Muse models, leaving analysts and startup executives saying it must rebuild trust and prove it can sustain an ecosystem.

Insights

Will Meta and Nvidia's aggressive open-source pivot successfully reclaim developer loyalty from dominant, ultra-cheap Chinese AI rivals?
Can Meta's turbulent AI division truly rebuild enterprise trust after shifting away from its open-source roots last year?
Does releasing advanced AI blueprints protect American tech leadership, or inadvertently hand foreign adversaries the exact tools they need?