SpaceXAI Launches Grok 4.7 for Coding, Long-Horizon Agentic Knowledge Work

SpaceXAI has launched Grok 4.7, its latest AI model focused on coding and professional knowledge work, claiming improvements in speed, reasoning, and handling complex tasks.

Grok 4.7 uses a larger base model and longer reinforcement-learning training to improve its ability to work on tasks that take hours to complete. According to xAI, the model performs better at checking its own work, creating documents and presentations, and handling cybersecurity tasks. 

It is available through Grok Build, Cursor, and the Grok API, starting at $2 per million input tokens and $6 per million output tokens.

SpaceXAI said Grok 4.7 is designed for coding and professional knowledge work. It uses a larger base model than Grok 4.6. The model underwent a longer reinforcement-learning training process focused on more difficult tasks, particularly those that can take several hours to complete.

“Grok 4.7 is a strong combination of intelligence, speed, and low cost,” Elon Musk, Founder of SpaceXAI (formerly xAI), stated on X.

The company claims, for biological safety, Grok 4.7 scored 62.4% on LatchBio’s biosafety benchmark. 

On the AA Briefcase benchmark, which evaluates long-horizon agentic work, Grok 4.7 scored 1,657, compared with 1,546 for Grok 4.6. On the Artificial Analysis Coding Agent Index, which evaluates end-to-end software engineering performance, it ranks fourth, only behind Claude Fable 5.1, GPT-6 Astra, and Claude Opus 5.

The model comes with a new safeguard system focused on cybersecurity and other dual-use areas. Grok 4.7 demonstrated stronger resistance to jailbreak attempts while maintaining its ability to assist with legitimate security work, according to the company. It reported that 3.3% of risky dual-use prompts were allowed through on its HackerBench v0.3 evaluation.

The company has also begun providing selected cybersecurity partners with invite-only access to the model’s red-team capabilities for defensive research.