Back in June of this year, when OpenAI initially introduced the GPT-5.6 series models, it announced that the Sol variant would be made available on the Cerebras platform in July. This move was aimed at providing certain clients with a remarkable processing speed, reaching up to 750 tokens per second. Nevertheless, during July, OpenAI refrained from divulging specific operational particulars of Sol on the Cerebras platform. Following that, OpenAI staff verified on the X platform that the company is currently undertaking an internal preview of this high-speed mode with select clients.