Recently, Elon Musk announced on his social platform that the latest generation of large language model Grok 4.5 has started a private test in the internal environment of SpaceX and Tesla. This move indicates that the model is moving from the laboratory to complex real-world applications.

According to Musk, the technical foundation of Grok 4.5 is the V9 model with 1.5 trillion parameters. To further enhance its logical and processing capabilities, the development team specifically introduced data from the popular AI programming tool Cursor in subsequent training. This targeted training strategy clearly aims to improve the model's accuracy in code writing and handling complex tasks.

In terms of performance, the early evaluation data of Grok 4.5 is impressive. Tests show that the model's overall capabilities have approached, and in some dimensions, it may even surpass the current flagship product of Anthropic, Claude Opus. Currently, the development team is continuously optimizing the model through reinforcement learning (RL) and improving the accompanying "Grok Build" testing benchmark to ensure stability under various stress tests.

Musk also sent a clear signal: the pace of large model development at SpaceX will accelerate further. He stated that in the remaining time this year, SpaceX will maintain an extremely high iteration frequency of releasing a new model every month. All models will be trained from scratch. This intense technological evolution not only demonstrates Musk's continued commitment to the AI field but also signals that the competition in the large model sector is about to enter a more frequent "battlefield" era.