Musk: Grok5 may surpass all current AI large model products.

Beating AI News Flash: Musk posted on social media, saying: "Grok 4.7 should be roughly on par with Opus 5.0, not 5.1. It is better in some ways and weaker in others. We need to improve multimodal performance. Grok 4.8 will bring a noticeable improvement. Grok 4.9 may reach Astra/Fable level. Grok 5 might surpass all current products. We'll see."

Previously, Musk teased Grok 4.8. He said 4.8 is a 2.5T parameter-scale model, trained with SpaceXAI's new C++ software stack, with pretraining to finish this week, followed by reinforcement learning.

This C++ stack is also a pie in the sky that Musk has repeatedly drawn before. At the end of June, he said he wanted to substantially rewrite Grok's training and inference software in C/C++ within about three months, remove a large number of intermediate layers, and make dedicated optimizations for NVIDIA GB300, expecting a "huge improvement" in about three months.

However, Grok 4.7 was just delayed due to RL issues. Musk previously said it would be released around September 12, but near launch he admitted that the model would give up too early on difficult problems and was not strict enough in checking answers. Before the previous pie is even out of the oven, the next one is already in.