SpaceX AI's Grok 4.7 is its most advanced model yet
SpaceX AI just rolled out Grok 4.7, its most advanced model yet for coding and knowledge work.
It's built to handle tougher, longer tasks with better speed and lower costs than before.
Plus, it comes packed with stronger safety checks and smarter task verification.
Grok 4.7 trained on tougher problems
Grok 4.7 was trained on more challenging problems and now gets the hang of conversations and document creation even better, thanks to its more expansive base model and native support for the Grok Bot harness.
On the security side, only 3.3% of potentially dangerous dual-use prompts were restricted on the HackerBench v0.3 benchmark (so you can do more without hitting a wall).
You can use Grok 4.7 via Cursor, Grok Build, or API for $2 per million input tokens ($6 per million output), with a speedier version available if you need things done even faster.