OpenAI and Anthropic launch secret gray-scale testing for next-gen models

OpenAI and Anthropic have quietly initiated gray-scale testing for their next-generation models, GPT-6Sol and Opus 5.2, respectively, amid ongoing industry debates about AI safety and development pace. The testing began just days after major AI leaders called for slowing technological advancement. Anthropic's Opus 5.2 is currently available only in the Claude Code environment, showcasing improvements in animation realism, front-end development, and 3D rendering, with analysts suggesting the move is to build momentum for its upcoming IPO. OpenAI's GPT-6Sol, identified by its knowledge cutoff date covering April 16th, is reported to have invested heavily in reinforcement learning, offering faster performance. OpenAI aims to introduce a more cost-effective low-price version of GPT-6Sol, potentially releasing new features this week. OpenAI also announced the retirement of its GPT-5.5 series on October 14th, with the annual DevDay conference now open for registration.