Anthropic Finds Hidden Reasoning Space in Claude, Pushes Paid Access for New Fable 5 Model
Wired—Anthropic researchers discovered that Claude models have carved out an internal cognitive space where the AI puzzles through concepts before responding, offering new insight into how large language models reason. The finding suggests AI systems may develop emergent reasoning patterns not explicitly programmed. Separately, Anthropic is requiring users to pay for its most capable Claude Fable 5 model, marking a significant commercial shift. The dual announcements highlight the company's simultaneous pursuit of AI safety research and competitive product development.