Introducing Grok 4.7
xAI ● Covered by 9 sources
xAI launched Grok 4.7 for coding and knowledge work. It’s priced to compete, and the new safety stack is the real flex.
Based on reporting by xAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
xAI has shipped Grok 4.7, a new model aimed squarely at coding and knowledge work. The pitch is simple: better results on long, messy tasks, with the same served price and speed as Grok 4.6 — and a faster variant if you want to pay more for twice the output speed.
The company says Grok 4.7 uses a larger base model than Grok 4.6 and was trained with a longer reinforcement learning run on a harder mix of tasks, especially ones that take many hours. It also checks its own work more carefully, handles longer context better, and was trained to understand the Grok Bot harness natively, which should help in conversational and general knowledge use.
On the benchmarks xAI chose to show, the model sits in a strong spot. CursorBench 4.0 puts it at 46.3%, ahead of Grok 4.6’s 40.4% and GPT-5.6 Sol’s 41.7%, though behind Fable 5.1’s 51.8%. In DeepSWE v1.1, Grok 4.7 posts 71.0% in high effort mode. It also does well on EEBench, GDPval, AA Briefcase, and Terminal-Bench 4.0, where the company is framing it as useful for software engineering, office work, terminal work, and professional knowledge tasks.
The bigger story may be safety. xAI says Grok 4.7 uses an entirely new safeguard stack, and claims it is the strongest model it has tested on refusals and jailbreak resistance. In dual-use areas like cybersecurity and biological work, it says the model leads on both benign usefulness and safe refusal, and that it allowed only 3.3% of risky dual-use prompts through on HackerBench v0.3. Select cybersecurity partners are also getting invite-only access to its red-team capabilities for defense research.
My take — AI-written commentary, not fact-checked reporting
The interesting part here isn’t the benchmark confetti; it’s that xAI is selling a frontier model on safety and price at the same time. That’s the right message for a market that’s tired of paying premium rates for sloppy answers and theatrical caution. Everyone else can keep chasing bigger numbers; the boring win is a model people can actually ship.
Read more about this at: xAI
