Artificial intelligence company Anthropic, which is racing toward an IPO, has released two frontier models built on identical underlying capabilities—Claude Fable 5.2 and Claude Mythos 5.2—differing only in their safety safeguards and access permissions. Compared with the previous generation, the new models are faster and more competitively priced, deliver outstanding performance in areas such as complex software engineering and scientific research reasoning, outperform contemporaneous frontier models from OpenAI and Google on multiple evaluation benchmarks, and feature an improved safety alignment mechanism. Anthropic is about to kick off its IPO roadshow, with current market valuation expectations having been revised downward multiple times from earlier levels. The company faces pressures including soaring computing costs, an operating loss in the third quarter, and massive long-term commitments for computing power procurement and leasing. Releasing its most powerful model ahead of the roadshow serves to demonstrate to investors both its technological iteration capability and its differentiated competitiveness—while also raising market concerns that stronger models will further intensify cash flow pressure.