SpaceXAI introduced Grok 4.7 on September 21, describing it as its most capable model for coding and knowledge work. The company says it checks its work more carefully and includes its most developed safeguard stack to date.
The standard model is served at the same price and speed as Grok 4.6: starting at $2 per million input tokens and $6 per million output tokens. The published CursorBench 4.0 comparison emphasizes the cost of completing longer coding tasks.
The comparison still places Fable 5.1 ahead on that benchmark, but Grok can be competitive when the budget for reaching a useful result matters. The chart in the original post also compares its task costs with GPT-5.6 Sol.
My view is that Grok 4.7 improves on its predecessor and has a credible price argument. That does not settle which model will perform best on a particular codebase. Test it on the kind of work you actually need done and count the cost of corrections as well as the first response.
Official announcement: Introducing Grok 4.7.
Adapted from the original Chinese article, published on September 22, 2026.
Comments NOTHING