SpaceXAI Releases Grok 4.7 for Coding and Knowledge Work

SpaceXAI has introduced Grok 4.7, its newest base model built for coding and knowledge work. The release features a larger base model trained with an extended reinforcement learning run focused on multi-hour tasks, better self-verification, and context management.

On benchmark evaluations, Grok 4.7 performs at the frontier for price-performance on CursorBench 4.0, which stresses long-running coding tasks. The model also improves on GDPval and AA Briefcase tasks previously handled by professionals such as nurses, lawyers, and financial analysts.

The model incorporates an entirely new safeguard stack. It scored 62.4% on LatchBio's biosafety benchmark. On HackerBench v0.3, the system allowed 3.3% of risky dual-use prompts through while maintaining utility for legitimate security work. Select cybersecurity partners receive invite-only access to red-team capabilities for defense research.

Pricing starts at $2 per million input tokens and $6 per million output tokens. A fast variant offering twice the output speed is available at twice the price. Grok 4.7 is accessible in Cursor, Grok Build, the Grok API, third-party coding harnesses, model routers, and cloud platforms.