Elon Musk teases Grok 4.6 and 4.7 as AI race heats up

Elon Musk on Tuesday signalled that SpaceXAI's aggressive model-release cycle will continue, posting on his social media platform X that Grok 4.6 and 4. are on the way.

He said Grok 4.6 will arrive around August 7 as a 1.5-trillion-parameter model with substantially upgraded supervised fine-tuning and reinforcement learning. This will be followed just weeks later by Grok 4.7, a larger 2.1-trillion-parameter model.

"This will be better than 4.6 in every way, except slightly slower to serve, albeit with even better token efficiency," he wrote.



The announcement builds on the momentum of Grok 4.5, which launched to the public on July 8, after Musk said strong positive feedback from customers in the beta test program had convinced the company the model was ready for wider release.

Musk had earlier described that model as an "Opus-class" model, but faster, more token-efficient and lower cost. He also posted later that internal evaluatoins placed Grok 4.5 model comparable to, and in some cases better than, Anthropic's Opus 4.7.

Grok 4.5 was built on a 1.5-trillion-parameter V9 foundation model, with supplemental training on data from Cursor. After launch, Grok drew a mixed but largely positive reception.

Grok 4.5 is ⁠priced at $2 per million input tokens and $6 per million output tokens. ‌Rival Anthropic's Claude Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens. Comparatively, OpenAI's GPT-5.6 Luna is priced at $1 per ⁠million input tokens and $6 per million output tokens.

The new Grok releses are positioning xAI right in the middle of the competitive AI space, particularly against rivals like Moonshot's Kimi K3 and Anthropic's newly released Opus 5.