Full Summary
This Sunday morning, Naver is doubling down on its AI search, integrating existing services like shopping and reservations into a powerful new "AI Tab." Both Aju Press and 아시아경제 confirm this new lightweight model, optimized for conversational AI, is twice as fast and reduces "hallucinations" by 30% compared to its predecessor. 매일경제 adds that this "Product Native Giant Languages Model" is trained with "AGENTIC AI," designed to complete user tasks by linking directly with Naver's services. Meanwhile, a major shake-up in the global AI landscape comes from China. Zhipu AI has released GLM-5.2, an open-weight AI coding agent. Tech My Money, Pragativadi, and ai-checker.webcoda.com.au all report this model features a 1-million-token context window for long software tasks and is available under an MIT license, making its weights downloadable. The Express Tribune highlights that GLM-5.2 rivals top Western systems, including Anthropic's Claude Opus, but costs significantly less—about one-quarter the price. This release follows the US government's decision, reported by Memeburn, to partially reverse a ban on Anthropic's Mythos 5, allowing limited access to trusted US organizations due to national security concerns. Microsoft CEO Satya Nadella, according to AOL.com, emphasizes that every company needs to build its own AI model tailored to its unique business needs, warning against the concentration of AI capabilities. This sentiment resonates with the emergence of open-source, lower-cost models like GLM-5.2. This means businesses now have more affordable, high-performance AI options, potentially changing their development costs and increasing competition in the AI market.