Alibaba's QwQ-32B challenges top models, Andrew Ng releases universal LLM framework, NVIDIA speeds inference, Deepseek cuts costs, and LLMs master neuroscience predictions.