Linked and quoted from the labs, the press and the people using them. Times in UTC.
The 17 labs we track and the biggest stories beyond them. The rest is in Everything.
DeepSeek published the DeepSeek-V4-Flash-0731 model on Hugging Face under an MIT license. The model card lists text-generation, transformers, and safetensors tags.
Google released Gemini Robotics ER 2, a model for robotic applications. It adds video understanding, tool orchestration and multi-robot collaboration.
Google launched Lyria 3.5 in Google Flow Music. The update brings advances in musicality, lyrics, vocals, and creative control.
Google DeepMind announced Gemini Robotics 2, a model that brings whole body intelligence to robots. The announcement was published on deepmind.google.
MiniMax released MiniMax-H3 on Hugging Face, a model supporting text-to-video, image-to-video, image-text-to-video and video-to-video generation. The repository lists 3,664,216 downloads and 5,623 likes.
Microsoft introduced MAI-Cyber-1-Flash, a compact security model that scores 96 percent on the CyberGym benchmark within its MDASH multi-agent system. Microsoft says costs drop about 50 percent versus frontier models since only tough cases go to GPT-5.4.
Microsoft released a Fantasy Premier League Companion tool for managers. The post announcing it appeared on Microsoft's Source blog.
Cohere introduced North Automations, a feature for intelligent workflow orchestration. The announcement was published on Cohere's official site.
Anthropic's Claude Opus 5 scored 30.2 percent on ARC-AGI-3, nearly quadrupling GPT-5.6 Sol's previous record of 7.8 percent. The benchmark's developers say the model independently formulated reflection equations, a behavior they had not seen from another model.