Amid intense discussions in the AI industry about safety and development pace, and just days after the leaders of two major giants publicly called for slowing down the rapid advancement of technology, the rate of internal competition within the industry has not slowed down at all. Just after Monday this week, OpenAI and Anthropic quietly initiated gray-scale testing for their next-generation flagship models in a coordinated manner.

According to feedback from multiple netizens and technical communities, OpenAI is currently conducting a gray-scale test of its new GPT-6Sol, while Anthropic skipped the 5.1 version and directly launched the highly anticipated Opus5.2. The fact that these top-tier companies are simultaneously advancing new model tests at the same time indicates different strategic considerations and market strategies behind each.

In terms of specific product updates, Anthropic's Opus5.2 is currently only available in the Claude Code environment, while the web version remains mostly on the 5.0 version. Early testers found that compared to previous versions, Opus5.2 has made significant breakthroughs in the realism of animation presentation, front-end development, and 3D rendering and building capabilities. Its actual performance is already capable of competing head-on with GPT-6Astra. Industry analysts believe that Anthropic's accelerated push for new model testing is mainly to build momentum for its upcoming IPO (Initial Public Offering).

On the other hand, the gray-scale signs of OpenAI's GPT-6Sol can primarily be identified by the model's knowledge cutoff date. If the model can accurately answer the competitive dynamics released on April 16th, it usually means it has been successfully routed to the test channel. Some early users reported that GPT-6Sol has invested a great deal of resources in reinforcement learning, and its overall operation speed is extremely fast. It is reported that one of the core purposes of OpenAI's increased efforts in promoting GPT-6Sol is to introduce a more cost-effective low-price version, thus reviving the long-rumored $200 premium subscription plan. New features may be officially released this week.

Along with the new model testing, official old product iterations are also being advanced simultaneously. OpenAI has officially announced that the existing GPT-5.5 series will be retired on October 14th, making way for the new generation of models. At the same time, the annual OpenAI DevDay developer conference has already opened registration, and it is expected to bring more quota benefits and technical ecosystem support to developers. These two leading AI companies are showing caution in words, but the competition in the product field is becoming increasingly intense.