AI news, analysis, and tools — with a focus on China's role in the global AI race
NVIDIA launched Nemotron 3.5 Lightning — a 30B MoE model with 3B active parameters, designed for high-throughput agent execution. Combined with NeMo Switchyard, it slashes costs by 74% while maintaining frontier-level performance.
NVIDIA released Nemotron 3 Embed — a family of open-weight text embedding models. The 8B flagship claimed #1 on RTEB with 78.5%. A compact 1B variant delivers 2x throughput on Blackwell GPUs. Why this matters for the RAG and AI agent ecosystem.
DeepSeek V4 officially launches with peak-valley API pricing. V4-Pro beats GPT-4o on coding benchmarks, offers 1M context for free, and is still 6x cheaper than GPT-4o even at peak rates.
China's DeepSeek disrupts the global AI industry with massive cost advantage.
Moonshot AI launches open-source code model built on trillion-parameter MoE.
Apple's AI strategy shift has implications for the China-US AI race.
On-device AI processing doubles — the mobile AI race heats up.
China's scientific AI reaches global #1 in molecular simulation.
Nvidia-MediaTek partnership powers next-gen AI hardware.