According to OpenClaw on X, The ClawCast Episode 14 highlights NVIDIA RTX Spark as an approach for running local AI on RTX GPUs, signaling growing on-device inference interest.

According to DeepSeekHarness, packaged desktop apps for macOS and Windows are live, with Linux via npm @ deepseek-ai/dsh, enabling local AI workflows.

According to @emollick, MiniMax H3 generated an otter-on-airplane video with sound locally in ~3 minutes, showing rapid open weights progress.