SpaceXAI Unveils The Grok 4.7 AI Model
Credit: XAI
SpaceXAI has released a new version of its AI model, Grok 4.7. The company bills it as the most advanced model for programming and intelligent work, with execution taking only a few hours.
In the CursorBench 4.0 test, which focuses on long-term coding tasks, Grok 4.7 demonstrated market-leading price-to-performance value. Compared to its predecessor, the new model underwent more extensive reinforcement learning on a more complex set of tasks, specifically targeting those requiring hours to complete. Grok 4.7 excels at verifying its own output and managing longer context windows. The model was trained to natively support the Grok chatbot framework, enabling it to handle conversational tasks and general knowledge queries more effectively.
SpaceXAI’s Grok 4.7 has improved its ability to generate high-quality office documents and presentations. In the GDPval and AA Briefcase tests—which involve performing professional tasks in fields such as law, financial analysis, and caregiving—the new model outperformed its predecessor and matched the performance of other leading systems. It features an updated security suite, enabling Grok 4.7 to achieve the best metrics in the family regarding the refusal of malicious requests and resilience against hacking attempts. It excelled in handling dual-use queries in cybersecurity and biotechnology, proving helpful for benign requests while refusing dangerous ones; it took first place in the LatchBio test with a score of 62.4%.
The main advantage of Grok 4.7, however, is its price: $2 per million input tokens and $6 per million output tokens—significantly lower than that of its competitors. Grok 4.7 is already available via the Grok API, Cursor, and Grok Build.
