SenseNova 5.0

Major upgrade of SenseTime's flagship large model family, launched at SenseTime Tech Day in Shanghai. Trained on over 10TB of tokens with synthetic data, adopts Mixture of Experts architecture with ~200K effective context window. Claimed to outperform GPT-4 Turbo on multiple benchmarks including MMBench for multimodal evaluation. Introduced the "Cloud-to-Edge" full-stack deployment matrix.