OpenAI's $50B Amazon deal clears

zdnet.com

Third-party AI assistants now work in Apple CarPlay, with ChatGPT winning out over Perplexity for conversational engagement, while Perplexity leads on location-based tasks—expanding the competitive landscape for in-car AI beyond Siri.

PyTorch Blog

SMG solves the Python GIL bottleneck in LLM serving by disaggregating all CPU workloads (tokenization, tool orchestration, multimodal processing) into a pure Rust gRPC gateway, delivering up to 3.5x throughput gains in production scenarios. The project has been adopted by major cloud providers and integrated upstream into vLLM and TensorRT-LLM.

Apple Machine Learning Research

Apple researchers propose LaDiR, combining latent diffusion models with LLMs to enable iterative refinement and parallel exploration of reasoning paths, outperforming traditional autoregressive chain-of-thought methods on mathematical and planning tasks. This approach addresses limitations of sequential token generation by allowing models to revise reasoning holistically.

Nvidia Blog

Nvidia's new Nemotron-3 Nano Omni consolidates multimodal AI capabilities into one efficient open model, enabling faster agents that process video, audio, and text simultaneously without context loss or latency penalties.

Reuters

OpenAI's $50B Amazon compute deal moves forward with revised terms that preserve Microsoft's revenue share, reflecting the company's shift toward multi-cloud infrastructure strategy despite existing partnerships.