xAI, the company behind the Grok series of artificial intelligence models, has launched a new version called Grok 4.7, which offers improved performance in coding, professional tasks, and security. This update includes a larger base model and enhanced training that focuses on complex, time-consuming tasks. The model is available through various platforms, including Cursor, Grok Build, the Grok API, and cloud services. Grok 4.7 maintains the same pricing as its predecessor, Grok 4.6, which starts at $2 per million input tokens and $6 per million output tokens. In coding and engineering tasks, Grok 4.7 performs better than previous versions. On the CursorBench 4.0 test, which evaluates long-duration coding challenges, it scored 46.3%, outperforming Grok 4.6 (40.4%) and GPT-5.6 Sol (41.7%). It also scored 71.0% on DeepSWE v1.1, a test that evaluates high-level reasoning, slightly behind GPT-5.6 Sol (72.7%) but ahead of Fable 5.1 (70.0%). On EEBench, which tests electrical engineering skills, Grok 4.7 achieved the highest score among compared models at 64.0%. The new model also shows improvements in handling professional tasks. On AA Briefcase v1.1, which evaluates multi-hour office tasks, Grok 4.7 scored 1,657 points, surpassing Grok 4.6 (1,546) and GPT-5.6 Sol (1,487). On the Harvey Legal Agent Benchmark, which tests legal skills, Grok 4.7 achieved the best score among the compared models at 19.6%. The model also performs well in creating documents and presentations and outperforms previous versions on GDPval and AA Briefcase, two benchmarks that assess the skills of professionals like lawyers, nurses, and financial analysts. xAI has also introduced a new security mechanism in Grok 4.7, which improves its ability to reject dangerous requests and resist attempts to bypass its safety protocols. Internal evaluations show that it performs better than any previous model in this area. On the LatchBio biosafety scale, it scored 62.4%, and on HackerBench v0.3, it accepted only 3.3% of risky prompts that could be used for harmful purposes. Some cybersecurity partners invited by xAI can access "red team" features for defensive research. Despite these improvements, the pricing for Grok 4.7 remains unchanged from the previous version, with a faster version available at double the cost.