Grok 4.7 Coding & Knowledge Work Model
Released by SpaceXAI on 21 September 2026 and billed as its most powerful coding and knowledge-work model: a larger base model with longer RL on multi-hour tasks, at the same price as Grok 4.6
Model Overview
Specs, access and pricing details below
Features
- ✓ Built on a larger base model than Grok 4.6 and trained with a longer reinforcement-learning run on a harder mix of tasks weighted toward problems that take many hours, better at verifying its own work and managing long context
- ✓ Natively understands the Grok Bot harness, making it stronger at conversational tasks and general knowledge work
- ✓ Same price and speed as Grok 4.6: $2 per million input tokens and $6 per million output tokens
- ✓ A new safeguard stack that SpaceXAI calls its strongest model yet on refusals and jailbreak resistance, allowing only 3.3% of risky dual-use prompts through on HackerBench v0.3 while rarely blocking legitimate security work
- ✓ Available at launch in Cursor and Grok Build, and through the Grok API and third-party coding harnesses
How to Use
- Switch to and select Grok 4.7 in Cursor or Grok Build
- Or use the Grok API, set the model to grok-4.7 and explicitly set reasoning.effort to high / xhigh when you want that ceiling
- Submit multi-hour coding or knowledge-work tasks and observe its self-verification and long-context handling
- Compare token usage and average cost per task (for example on CursorBench 4.0) against Grok 4.6
FAQ
What is Grok 4.7?
A model released by SpaceXAI on 21 September 2026 and billed as its most powerful model for coding and knowledge work, using a larger base model than Grok 4.6 and a longer reinforcement-learning run on multi-hour tasks.
How does it differ from Grok 4.6?
SpaceXAI says Grok 4.7 works longer on difficult tasks, checks its own work more carefully and manages longer context better, and it natively understands the Grok Bot harness, all at the same price and speed.
How much does Grok 4.7 cost?
$2 per million input tokens and $6 per million output tokens, the same as Grok 4.6.
Where can I use it?
It is available at launch in Cursor and Grok Build, and through the Grok API and third-party coding harnesses.
How does Grok 4.7 perform on safety?
SpaceXAI says it uses an entirely new safeguard stack and is the strongest model it has tested on refusals and jailbreak resistance, allowing only 3.3% of risky dual-use prompts through on HackerBench v0.3 while rarely blocking legitimate security work.
How do I control the thinking effort?
The Grok API defaults to high, and several benchmarks in the model card are reported at xhigh; set reasoning.effort explicitly when you want that ceiling.