NVIDIA Cosmos 3: Open Physical AI Model That Predicts Robot Actions
NVIDIA Cosmos 3 is now on Hugging Face — a 16B/64B open foundation model that generates robot trajectories, not just video. Here's what ...
MiniMax M3: Frontier Coding Model at 5% of Claude’s Price
MiniMax M3 launches today with 59% SWE-Bench Pro, 1M-token context, and API pricing at $0.60/M ...
AV2 Video Codec Drops: 30% Better Than AV1, dav2d Now Live
AV2 video codec released with dav2d decoder from VideoLAN. Expect 30% better compression than AV1, ...
Apple’s Container CLI: Run Linux on macOS Without Docker Desktop
Apple's container CLI runs OCI Linux containers in per-VM isolation on macOS 26. Here's when ...
Hermes Agent: Self-Improving AI Runs Locally on RTX
Hermes Agent hit number one on OpenRouter with 224 billion daily tokens. Here is how ...
AV2 Codec Is Finalized: dav2d Ships and the 40% Compression Gap
AOMedia finalized AV2 on May 28 and VideoLAN shipped dav2d — the first open-source AV2 decoder. Here's what the 40% compression gain means ...
Apache Kafka 4.3: Broker Cordoning, CIDR ACLs, Share Groups
Apache Kafka 4.3.0 ships broker cordoning, CIDR-based ACLs, and share group tuning. Here's what each ...
KernelScript 0.1: A Type-Safe DSL for eBPF Development
KernelScript 0.1 ships a type-safe DSL that generates eBPF kernel code, userspace loaders, and Makefiles ...
Stable Audio 3.0: Open-Weight AI Audio You Can Actually Ship
Stable Audio 3.0 ships open weights, LoRA fine-tuning, and CPU inference. Here's the developer guide ...
DeepSeek V4: MIT Open-Source Model Matches Claude and GPT-5.5
DeepSeek V4-Pro matches Claude Opus 4.7 on SWE-bench at $1.74/M tokens under an MIT license. ...