Elon Musk’s SpaceXAI unit proposes an artificial intelligence model Glock 4.7 – It is targeted primarily for writing code and scientific tasks. The developers stress that the speed and cost of use remain unchanged compared to the previous Grok 4.6 – this model is very competitive in its class.
Photo credit: Rubaitul Azad/unsplash.com
In the CursorBench 4.0 test, which focuses on long-term coding tasks, Grok 4.7 proved to be the leading product in terms of price/performance. The new model is trained with longer reinforcements on more complex problems than the previous model, focusing on those that require a significant amount of time to solve. Grok 4.7 does a better job of inspecting its own output and managing longer context. The model is trained at the native level to support the Grok chatbot software framework – which can better handle conversational format tasks and general knowledge requests.
The following image source: x.ai
SpaceXAI Grok 4.7 Start building better office documents and presentations. In GDPval and AA Briefcase tests covering professional tasks in areas such as law, financial analysis and personal care, the new model outperformed its predecessor and was on par with other leading systems. It features an updated security toolset – as a result, Grok 4.7 demonstrates the best performance in the series when it comes to refusing to respond to malicious requests and fending off hacker attacks. It leads the way in dual-use requests in the cybersecurity and biotech fields, helping with benign requests and rejecting dangerous ones—it ranked first in the LatchBio test with a score of 62.4%.
Another advantage of this model is its good network security behavior: in HackerBench v0.3, it allowed only 3.3% of risky dual-use requests and rarely blocked legitimate security requests. SpaceXAI has provided Grok 4.7 to some partners as a stress testing tool for security development.
The new model is already available in Cursor and Grok Build through the Grok API, third-party code generation services and cloud platforms. Price is $2 for 1 million input tokens and $6 for 1 million output tokens; options for double the speed and price are available.
If you find an error, select it with your mouse and press CTRL+ENTER.
