join.dev
AI news · 10 stories
Google DeepMind previews Genie 4 for interactive world simulation · Liquid AI releases LFM3 for specialized multimodal models · NVIDIA expands its Meta partnership and plans $3 billion investment
Google DeepMind announced Genie 4, adding native audio and real-time language control to interactive world simulation. Access starts with trusted testers, with potential uses in robot training and interactive prototyping.
Read at Google DeepMind
Liquid AI introduced LFM3, a family of compact text, vision, audio, omni and decision models sharing a common backbone. The models support single-GPU customization and deployment on local hardware, giving developers more options for specialized applications.
Read at Hugging Face
NVIDIA and Meta announced expanded work on Muse training and inference, alongside plans for broader NVIDIA infrastructure deployment. NVIDIA will invest $3 billion in Meta’s equity offering, linking the technical partnership with additional capital.
Read at NVIDIA
Microsoft launched MAI Image 2.7 in preview for image editing through Arena, Artificial Analysis and MAI Playground. The release targets scene restyling, relighting and detailed asset edits, with text-to-image generation planned later.
Read at Microsoft AI
Oracle announced an expanded collaboration combining OpenAI models, ChatGPT and Codex with its applications, data and cloud infrastructure. The companies aim to bring AI into enterprise workflows, application development and security work.
Read at Oracle
General Compute announced a $100 million Series A co-led by Blackbird and Abstract to expand its inference cloud across GPUs and specialized chips. It plans its first major deployments for Q1 2027, offering capacity without customers having to procure and operate the hardware.
Read at General Compute
Luxor announced a $16 million round led by CMT Digital, including both new investment and sales of existing shares. The mining-infrastructure company plans to expand AI hardware, cloud services and compute trading, extending its business into AI capacity.
Read at The Block
NVIDIA published a guide to CUDA 13.4 GPU fractioning with KAI Scheduler and the GPU Fractioning Operator. It shows how workloads can share GPU capacity with memory allocations and limits, helping teams use available hardware more efficiently.
Read at NVIDIA
Microsoft and Cyberlite launched an expanded AI Classroom Toolkit during Global Media and Information Literacy Week. The teaching resource focuses on responsible generative AI use and judging information, supporting educators as AI changes how students learn.
Read at Microsoft
Google and the Stanford Accelerator for Learning introduced Polar Peaks, an AI Quests lesson for students aged 11–14 inspired by brain-mapping research. Data-cleaning and model-testing activities, alongside new teacher modules, emphasize oversight and critical thinking.
Read at Google