SpaceXAI launches Grok 4.7 with double the speed at half the price

SpaceXAI released its new Grok 4.7 model on September 22, focusing on coding and knowledge work. The company says it uses a larger base model trained with longer reinforcement learning, making it its strongest coding and knowledge model yet — twice as fast as similar models at half the price.

Grok 4.7 performs well on long-running coding tasks in the CursorBench 4.0 test, ranking among the top in price and performance. It was also natively trained for Grok Bot harness, improving dialogue, knowledge work, and document and presentation generation. On GDPval and AA Briefcase benchmarks, it reached levels comparable to other frontier models.

For safety, Grok 4.7 uses a new protection system, achieving the company's highest level in refusal and jailbreak resistance. It is available through Cursor and Grok Build, with API access starting at $2 per million input tokens and $6 per million output tokens.