xAI releases Grok 4.7 with bigger context and better coding scores

xAI just dropped Grok 4.7, boosting performance on long coding tasks without raising token prices.

xAI announced Grok 4.7, its latest model for software engineering and complex knowledge work. The company built a larger base model and put it through a longer reinforcement learning run focused on multi-hour tasks. On CursorBench 4.0, Grok 4.7 hit 46.3%, beating GPT-5.6 Sol Max while keeping the exact same price as Grok 4.6.

Why it matters: You get a sharper model on long-running tasks for $2 per million input tokens and $6 for outputs. It also tops benchmarks in electrical engineering and legal work, while cutting dangerous prompt leak rates on cyber defense tests down to 3.3%.

Try it today in Cursor, through the Grok API, or test it for free in Grok Build at x.ai/build.

A smarter model at the same old price point is always a welcome update.