Grok 4.7 shipped today after the long cook delay. Same price and speed as 4.6 on the standard API. Bigger base model and more training on long jobs that take hours. Live in Cursor Grok Build and the API. GitHub Copilot is rolling it out too.
The catch people should read carefully. Standard rates match 4.6 but the model can burn a lot more output tokens when thinking is cranked up. Twice as fast at half the price is a rival comparison headline, not a cut to your old Grok bill. Consumer Grok inside X comes later per the model card. Coding and work tools are the start.
ByteForward walked through the demos benchmarks and the token catch here
https://www.youtube.com/watch?v=VKVwSA1MMIM
Top comments (0)