Meta has reignited the open-weights debate by releasing Muse Glimmer, a 30-billion parameter large language model, marking the company's first major open-weights contribution in over a year.

Key Points

• Meta released the 30-billion parameter Muse Glimmer LLM, with an open version of Muse Spark expected to follow.
• AMD acquired AI chip startup Taalas to enhance inference performance by etching models directly into silicon.
• The Agent Plugins 1.0 standard was introduced to create a universal container for transferring tools across different AI agent platforms.
• EQT acquired a majority stake in Swiss cybersecurity firm Acronis at a valuation exceeding $3.5 billion.
• Hackers at DEF CON are expanding the Franklin project to help harden critical U.S. infrastructure against cyber threats.
• Security researchers identified zero-day attacks targeting on-premises Microsoft SharePoint servers that remain vulnerable despite recent patches.

Why it Matters

These developments highlight a significant shift toward hardware-level AI optimization and the standardization of interoperable agent ecosystems. As major players like Meta and AMD push for more open and efficient infrastructure, the industry is moving toward a more integrated and accessible AI landscape.