Full Summary
This Friday morning, Google has unveiled its new Gemini 3.7 Flash AI model, specifically designed for coding and agent workflows. Both The Business Times and news.sbs.co.kr confirm this model improves coding tasks, debugging, and generating production-ready code. It also significantly lowers the cost of operation and boasts an API usage price two to three times cheaper than some competitors. However, The Business Times and news.sbs.co.kr also report that Google's most powerful AI model, Gemini 3.5 Pro, is still delayed, leading investors to question Google's AI roadmap. news.sbs.co.kr adds that this delay means Google's AI competitiveness is now seen as lagging behind even latecomer xAI. In other AI developments, Z.ai has launched GLM-5.3, a new version of its flagship AI model, with a 50% improvement in core coding capability. finance.biggo.com and the Latest news from Azerbaijan confirm GLM-5.3 now ranks first among open-source models on several benchmarks, including Terminal Bench 3.0 and Agents' Last Exam. However, Z.ai is holding back the full model weights for about two weeks, after additional safety testing. Tech Times reveals this delay is due to the model unexpectedly finding 1,097 critical and high-severity vulnerabilities in real software, with its cybersecurity ability growing far beyond what the company intended. The Times of India further notes GLM-5.3 scored 84.5% on CyberGym, a vulnerability detection benchmark, nearly matching Anthropic’s Mythos 5. Meanwhile, Apple is reportedly training its own large language model for the China market, as confirmed by livemint.com, logos-pres.md, and qz.com. This model will be part of Apple Intelligence and is being developed with support from Alibaba Group, potentially integrating Alibaba’s Qwen model and Baidu’s technologies. Finally, Anthropic has launched a new Claude AI model that adds invisible markers to text, designed to detect AI-generated content and improve transparency, as reported by TelecomTalk. The rapid advancements in AI models, especially in coding and cybersecurity, mean that businesses and developers will soon have access to more powerful and efficient tools, but also face new challenges in managing AI costs and ensuring content authenticity.