Elon Musk’s SpaceXAI has introduced the Grok 4.7 artificial intelligence model, aimed primarily at writing software code and solving scientific tasks. According to the company, the new release keeps the same operating speed and usage cost as the previous Grok 4.6 version.
SpaceXAI says Grok 4.7 delivered a leading result on price-to-performance in the CursorBench 4.0 test. The company added that the model went through a longer reinforcement learning stage and was trained for more complex assignments, including tasks that can take several hours to complete.
According to the developer, Grok 4.7 is better at checking its own output, handling long context and completing tasks in a conversational format. The model also produces higher-quality office documents and presentations.
In the GDPval and AA Briefcase professional benchmarks, the model outperformed the previous version and posted results that SpaceXAI says are comparable with those of other leading systems.
Safety was another focus of the update. The company says Grok 4.7 is more effective at rejecting harmful prompts and resisting attempts to break the system thanks to upgraded safety tools. In the LatchBio biotechnology safety test, it ranked first with a score of 62.4%, while in HackerBench v0.3 it allowed only 3.3% of risky dual-use prompts through, while rarely blocking legitimate defensive queries.
Grok 4.7 is already available in Cursor, Grok Build, through the Grok API, as well as in other code-generation services and cloud platforms. Pricing starts at $2 per 1 million input tokens and $6 per 1 million output tokens. A version with double the speed and double the price is also available.
