BlockBeats News, July 28, Musk tweeted that Grok 4.6 is expected to be released around August 7. This version will adopt a 1.5 trillion parameter model and achieve significant improvements in Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL).

Grok 4.7 is expected to be released in a few weeks, adopting a 2.1 trillion parameter model. This version will surpass Grok 4.6 in all aspects, the only drawback being a slightly slower inference service speed. However, its token utilization efficiency will be further improved.

Click on the original text link below to join the Block Beating · Feishu AI News Channel, which continuously monitors global AI hotspots and news 24/7.