Tencent open-sourced Hy4 preview, a 770B-parameter MoE model with 49B activated weights and a 1M-token context window. The Apache 2.0 release puts a frontier-cl...
New coverage from across AI, software, and hardware.
Tencent's Hunyuan team released Hy-MT2 on 21 May 2026 — a new translation model family with instruction-following templates, three sizes up to a 30B-A3B MoE variant, and open weights plus IFMTBench. This is the next generation after the April Hy-MT1.5 phone quantizations.
OpenAI, Anthropic, Microsoft, AWS and 112 other organizations warned Thursday that defenders have only months to prepare for AI-driven cyberattacks on critical infrastructure. The coalition urges governments and companies to fund and deploy AI-powered defenses before offensive capabilities outpace protection.
NVIDIA released TensorRT Model Connect in public preview, turning Hugging Face checkpoints into native C++ TensorRT bundles with two commands and no ONNX step. The aarch64-only wheels target robotics and edge teams on ARM-based NVIDIA hardware.
Anthropic opened a research preview of the Model Hardware Standard, a specification that lets AI agents safely control lab and factory equipment. The standard cuts device integration from months to minutes and is being tested by Genentech, AWS, and robotics vendors.
PyTorch Foundation published the Core PyTorch track for PyTorch Conference North America 2026 in San Jose on October 20–21. The lineup centers on torch.compile internals, cross-repo CI for out-of-tree backends, and distributed APIs infrastructure teams will ship against.
Google DeepMind released Gemini Omni 1.1 Flash on August 27, 2026, adding scene extension, keyframe control, and 4K upscaling for developers. The update makes generative video viable for professional creative workflows with controllable camera motion and cheaper iteration.
NVIDIA hand-delivered its first Vera CPUs to Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure in mid-May 2026, correcting earlier reports of late August shipments. The 88-core Olympus processor targets agentic AI workloads where CPU orchestration gates GPU utilization.
Tencent open-sourced Hy-MT1.5-1.8B packs that shrink a 3.3 GB translation model to as little as about 440 MB for offline use on ordinary phones. The 1.25-bit and 2-bit releases keep 33-language coverage while targeting on-device privacy and real-time translation without a network.
Anthropic detailed how Claude Cowork now syncs sessions across desktop, web, mobile, and the Chrome side panel using cloud execution. The rollout lets Pro, Max, and Team users start tasks on one device and finish them on another without keeping a computer awake.
