SpaceXAI founder Elon Musk posted on the social platform X on Tuesday local time that Grok 4.6 and Grok 4.7 are coming soon, meaning the company will continue maintaining a high-frequency release schedule for large language models.
Musk said Grok 4.6 is expected to launch around August 7, with a model size of 1.5 trillion parameters. It will receive significant improvements in Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL).
A few weeks later, SpaceXAI will also launch the larger Grok 4.7, whose parameter count will increase to 2.1 trillion, further enhancing the model's capabilities.
Grok 4.5 was officially made available to the public on July 8. Musk previously said that feedback from users participating in the beta testing program had been highly positive, leading the company to believe that the model was ready for release to a broader user base.
In an earlier introduction, Musk described Grok 4.5 as a model that had reached “Opus-level” performance, while emphasizing that it offered faster execution, greater Token utilization efficiency, and lower inference costs.
He later revealed that the company's internal evaluation results showed that Grok 4.5 had also achieved further improvements in multiple performance tests.