Part 12/16:
In China, SenseTime's SenseNova models and Tianjin-based startups like Minimax are claiming performance surpassing the US giants in multiple metrics, with models like SenseNova 5.5 outperforming GPT-4 in some benchmarks. Their open-source strategies and multimodal capabilities are potential game changers.
Anthropic’s Claude 3.5 Sonet is another standout, demonstrating rapid learning, superior reasoning, and improved understanding of humor and long-term context. It’s already beating models like GPT-4 on several benchmarks, and its flexible API support hints at widespread adoption.