This week in artificial intelligence was defined by model breakthroughs, massive funding injections, pivotal regulatory moves, and research leaps with real-world impact. Here are the stories that shaped the landscape.
New Models
OpenAI launched GPT-5, its most capable model yet, featuring cascade-of-thought reasoning that dramatically improves performance on math, law, and science benchmarks. GPT-5 natively handles text, images, audio, and video, and introduces an Agent Mode for autonomous task execution, with reduced hallucination rates reported.
Meta open-sourced Llama 4 in two sizes: a 70B-parameter version for local use and a 700B mixture-of-experts model for cloud deployment. Designed for agentic tasks, Llama 4 prioritizes efficiency and on-device capability, cementing Meta's commitment to open-weight AI.
Google DeepMind released Gemini 3, integrating advanced reasoning with real-time multilingual translation and multimodal understanding. It features a Truthfulness module to improve factual accuracy and excels in cross-modal applications like video analysis and robotics.
Funding
Anthropic closed a $5 billion Series D at a $100 billion valuation, with plans to scale compute infrastructure and safety teams. CoreWeave secured a $2 billion debt facility to expand AI-focused data centers.
In Europe, Mistral AI raised $600 million in Series C, rewarding its research-driven approach. In Asia, Baidu's AI chip spin-off KunlunTech secured $1.5 billion for next
Top comments (0)