Elon Musk: Grok 4.5 Enters Private Testing at SpaceX and Tesla
Elon Musk stated that Grok 4.5, based on the 1.5 trillion parameter V9 foundational model and supplemented with Cursor data, has entered private testing at SpaceX and Tesla.
Early evaluations show its performance is close to or even exceeds Opus, with reinforcement learning significantly enhancing the model, and the Grok Build tool improving daily.
In terms of market mechanisms, xAI's rapid iteration and internal deployment accelerate investments in AI infrastructure, with capital flowing into computing power, data labeling, and model training. The event-driven AI competition is intensifying, benefiting xAI and its associated ecosystem, while competitors lagging in iteration face pressure.
Source: Public Information
ABAB AI Insight
Elon Musk has previously accelerated the development of multiple generations of Grok models through xAI, utilizing real-world data and application scenarios from SpaceX and Tesla for closed-loop validation, continuously pushing AI from research to engineering deployment.
The capital pathway indicates that xAI and SpaceX are concentrating resources on the monthly release of new models, motivated by the need to maintain technological leadership through high-frequency iterations and provide differentiated intelligent capabilities for the group ecosystem, strategically building a super AI infrastructure that covers both the physical and digital worlds.
Similar to the internal iteration pathways of Tesla Dojo and FSD, xAI is currently in a rapid expansion phase from foundational models to industrial-grade monthly deliveries.
Essentially, this represents a technological replacement and capital concentration, with the mechanism being the combination of real-world data and reinforcement learning in a high-frequency release cycle, significantly compressing the window for competitors to catch up. Capital and talent are concentrating on teams with closed-loop ecosystems and monthly delivery capabilities, while pricing power is shifting to organizations that can continuously output cutting-edge models.
ABAB News · Cognitive Law
Foundational models lay the groundwork, real data builds the skyscrapers; closed-loop iteration determines AI height.
Monthly new models catch up for a moment, continuous delivery leads for a lifetime; rhythm is the super-intelligent barrier.
Retail investors chase concepts for a moment, institutions use internal Betas for a lifetime, top capital builds a monthly iterative AI empire structure.