x.ai Releases Grok 4.7 for Coding and Knowledge Work

Grok 4.7 is built on a new, larger base model compared to Grok 4.6. The training involved a longer reinforcement learning run on a harder mix of tasks, specifically weighted toward problems that take many hours to complete. These improvements enable the model to better verify its own work and manage longer context. Additionally, the model was trained to natively understand the Grok Bot harness, enhancing its performance in conversational tasks and general knowledge work.

In professional benchmarks such as GDPval and AA Briefcase, Grok 4.7 performs comparably to other frontier models. On CursorBench 4.0, the model demonstrated frontier price-performance for long-running coding tasks.

The model features an entirely new safeguard stack. On the LatchBio biosafety benchmark, it scored 62.4%, and on HackerBench v0.3, it allowed 3.3% of risky dual-use prompts through. x.ai reports the model leads in both utility for benign tasks and safe refusal of dangerous ones in domains such as cybersecurity and biological work.