4 results for “large language models”
Tencent open-sources AngelSpec, a unified training framework that uses workload-aware strategies — multi-token prediction for dialogue and block diffusion for code and math — to boost LLM inference throughput across concurrency levels 4-64.
Read more →
SpaceX has placed a $52 billion order with Foxconn for nearly 1 million Nvidia GB300 GPUs, marking one of the largest AI infrastructure investments. Deliveries are expected from Q4 2026 to Q1 2027.
Read more →
Apple Intelligence has been registered with China's cyberspace regulator, clearing a major hurdle for its launch in the country. Alibaba's Qwen model and Baidu's AI will power the Chinese version.
Read more →
NVIDIA has introduced the Jetson Thor T3000 and T2000 modules, giving robotics developers new options for running multimodal AI models and agent workloads directly at the edge.
Read more →