Overseas tech news confirms that OpenAI's flagship GPT-5.6 model (Sol, Terra, and Luna series) has received approval from the U.S. Department of Commerce and is scheduled for full public release this Thursday (July 9th). Meanwhile, SpaceX AI (integrating Cursor programming data) has announced the completion of its private beta testing for Grok 4.5, which will officially launch its public beta tomorrow. Coupled with the upcoming releases of Google Gemini 3.5 Pro and DeepSeek V4, global AI labs are being forced from quarterly updates to a weekly or even monthly upgrade cycle in an arms race. The major companies' rush to bring models to market is essentially an attempt to offset Wall Street's scrutiny of exorbitant CapEx prices with the output of terminal computing power before the Q2 earnings season.
Constellation Research and publicly available market data show that the competitive logic among cutting-edge models is undergoing a profound divergence: on one hand, OpenAI and Grok are pursuing extreme inference capabilities at the trillion/trillion level (1.5T+V9 architecture); on the other hand, DeepSeek V4 is rapidly capturing market share in the open-source market and agent scenarios thanks to its extremely high token cost-benefit ratio (e.g., as low as $0.09 per million tokens). This comparison has made the market realize that if high-cost flagship models cannot bring commensurate, monopolistic commercial returns, traffic will accelerate its shift towards inference devices and dedicated application-specific integrated circuits (ASICs) with extreme cost-effectiveness.
