Meta launched Muse Spark, its first model from the newly created Meta Superintelligence Labs. Meta claims Muse Spark marks the beginning of a "ground-up overhaul" of its AI strategy, which aims to fix the performance gap between the company's models and similar offerings from its competitors.
Arcee AI, a tiny US startup that recently developed a completely open contender to Meta's Llama 4 Maverick, recently released Trinity-Large-Thinking, an open weights reasoning model claimed to be "the strongest open model ever released outside of China."
Microsoft AI announced the availability of three foundational models for transcription and voice and image generation. The launch highlights these models' competitive pricing and practical business value as Microsoft shifts its AI strategy towards business and productivity solutions.
Mistral raised $830 million in debt financing from a consortium of seven banks to purchase 13,800 Nvidia GPUs for its first data center near Paris, positioning itself as Europe's sovereign AI alternative to US proprietary models.
Mercor recently confirmed it was victim to a security incident related to the compromise of the popular open-source project LiteLLM. Extortion group Lapsus$ has claimed responsibility for the attack, allegedly posting a sample of the stolen data on its leak site.
Wikipedia has implemented a new policy that bans AI-generated article creation and substantial rewrites. The ban on LLM usage was voted by Wikipedia's editor community, who found themselves increasingly overwhelmed by the non-stop flood of AI slop making its way into the encyclopedia.
Granola recently raised a $125M Series C at a $1.5B valuation. In addition to the round announcement, Granola also shared the three new features coming to the app: personal and enterprise APIs, an updated MCP, and Spaces for sharing and collaboration.
Interloom has raised a $16.5M seed round to develop a platform that captures undocumented operational expertise and transforms it into a permanent context layer for AI agents. With its "Context Graph", Interloom aims to address the critical knowledge gap that affects enterprise AI deployment.
Mistral has launched Forge, a platform that lets enterprises train AI models from scratch on their own proprietary data for greater accuracy and control.
Cursor recently released Composer 2, a new in-house coding model that vastly improves its predecessor's performance. While Composer 2's benchmark scores may not be outstanding, Cursor is betting that the model's lower price point and native integration to the coding environment will drive adoption.
Yann LeCun's AMI Labs raised a $1.03 billion seed round at a $3.5 billion valuation, Europe's largest seed round on record. The startup will use the raised money to continue developing world models that can be applied to robotics, industrial, and healthcare applications.
Encyclopedia Britannica and Merriam-Webster have sued OpenAI for copyright infringement and trademark infringement. The publishers accuse OpenAI of unlawful scraping and reproduction of their content and claim that falsely attributed hallucinations are damaging their reputations as trusted sources.