oQe models updated with improved quantization in oMLX 0.7.1

Apple researchers present Minions, a distributed virtual GPU cloud that pools idle GPUs across macOS devices to run large-scale ML inference without buying new hardware. The system adds fault tolerance and flexible task execution for teams facing constrained centralized GPU capacity.

graphsignal-top ranks GPU kernels by time, live, with readable names, inside the open-source Graphsignal profiler. It wraps vllm serve or sglang serve via graphsignal-run and serves measurements as JSON so an agent can read them in a loop.

Apple researchers tested LLMs against human annotators on MQM and Error Span Annotation across 70 language pairs plus WMT23 and WMT25 data. LLM agreement beat human-human agreement on some tasks but varied widely by scheme, language pair, and domain, staying unreliable in most settings.

Krea's official X account posted that Krea 3 is coming, with no release date, no confirmation of open-source availability, and no word on editing capabilities.

NVIDIA Developer livestream with partners RidgeRun and Connect covers reducing software overhead, choosing efficient inference runtimes, and quantizing models to fit larger AI models on lower-cost Jetson SKUs.