Full Summary
This Friday, May 15th, the AI landscape sees significant new model launches and intriguing developments. Both SMBtech and 디지털투데이 confirm Google DeepMind has released Gemma 4, an open-source AI model family designed for advanced reasoning. Google calls it their "most capable" open-source AI, with models ranging from 2B to 31B parameters, built on the same research as their proprietary Gemini 3. SMBtech reports developers have downloaded Gemma family models over 400 million times, with Gemma 4 outperforming rivals up to 20 times its size. Meanwhile, KuCoin and 디지털투데이 highlight the performance of advanced AI models in specific tasks. GPT-5.5, the current top performer, achieves only 52% accuracy on real-world financial analyst tasks, according to the Vals AI Finance Agent v2 benchmark, confirming AI cannot yet fully replace human analysts. However, a new project called AI IQ, reported by 디지털투데이, converts public benchmark scores into a human IQ scale, placing GPT-5.5 at the top, followed by GPT-5.4 and Gemini 3.1 Pro. Interestingly, Gemini models are noted as being more cost-effective for similar IQ scores. Infosecurity Magazine reports on new AI models impacting cybersecurity, with Anthropic's Mythos Preview and OpenAI's GPT-5.5-Cyber both capable of finding and fixing vulnerabilities. However, the concern is a potential "vulnpocalypse" as these tools could uncover too many zero-days for organizations to patch. In other news, GK Today confirms Anthropic and the Gates Foundation have launched a $200 million partnership to develop AI tools for global health, education, and agriculture, focusing on low- and middle-income countries. Separately, Pulse 2.0 announces Wondrlab's $100 million investment in WondrBridge, a joint venture creating AI-powered Global Capability Centres. Startup Fortune reveals InternLM's Intern-S2-Preview, a 35-billion parameter scientific AI challenging the need for trillion-scale models by achieving comparable results through task scaling. Finally, belsat.eu reports Belarus and Russia are developing a joint "ideologically correct" AI language model, while Startup Fortune notes the brief appearance and disappearance of a Microsoft "Lens" image model on Hugging Face, sparking developer speculation. This surge in AI development means both new tools and new challenges are emerging rapidly, impacting everything from global health initiatives to the security of your online data.