Elon Musk has confirmed that SpaceXAI is preparing to launch its latest AI model, Grok 4.6, around August 7. According to Musk, the upcoming model will feature 1.5 trillion parameters and incorporate significant improvements in Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL).
The announcement arrives shortly after the company’s rebrand from xAI to SpaceXAI in July 2026, marking a strategic integration with SpaceX’s broader operational infrastructure. Musk also teased the development of Grok 4.7, a more powerful 2.1 trillion parameter model, which he expects to release just a few weeks after Grok 4.6. Regarding the successor, Musk noted that it will surpass 4.6 in capabilities, offering improved token efficiency despite being slightly slower to serve.
The announcement follows a positive performance assessment from Vercel CEO Guillermo Rauch, who recently identified the existing Grok 4.5 model as a top-tier cybersecurity AI in terms of price-performance. Rauch reported that Grok 4.5 is significantly more cost-effective than competing models, including OpenAI’s Opus 5 and the Kimi K3, while maintaining competitive performance levels.
This rapid development cycle occurs as competition in the AI sector intensifies. Major U.S. firms, including SpaceXAI, OpenAI, and Anthropic, are currently navigating a competitive landscape that includes emerging international models. Meanwhile, industry leaders continue to debate the potential for federal AI regulations, with Musk and other executives recently opposing restrictions on open-weight AI models, arguing that such measures could hinder innovation.