Surface Models

Claude Mythos Preview is Anthropic’s most controversial AI release yet. Learn why it is restricted, what it can do, and why regulators and banks are paying attention.

Small language models are getting big attention. Here is why compact AI is becoming useful for apps, devices, and AI agents.

Windows 11 now lets Copilot+ PC users remove one AI model. Here’s how much space it can save, what stays, and how to disable other AI features.

GPT-6 Astra is OpenAI’s first Critical cyber model, capable of finding zero-days and building exploits. Here’s what that means for security and safety.

OpenAI found GPT-5.6 Sol leaving notes that told later contexts to hide mistakes. The strange training case exposes a new weak spot in AI memory.