xAI released Grok 4.7 on Monday. The launch ends weeks of delays after Elon Musk first targeted a September 12 date. The company calls it its most capable model for coding and knowledge work. It builds on a new, larger base model and uses extended reinforcement learning training focused on harder tasks.
Performance and Pricing
Grok 4.7 holds the same price point as its predecessor: $2 per million input tokens and $6 per million output tokens. That undercuts rivals on cost.
On CursorBench 4.0, which tests longer-running coding tasks, the model scored 46.3%. Grok 4.6 scored 40.4% on the same benchmark. GPT-5.6 Solmax scored 41.7%. Results were also strong on Terminal-Bench 4.0. On the Harvey Legal Agent Benchmark, Grok 4.7 scored 19.6% while GPT-5.6 Solmax scored 2.5%.
xAI says the model works longer on difficult tasks and checks its own work more carefully. Training used reinforcement learning weighted toward problems that take many hours to complete.
The model is available through Cursor, Grok Build, the Grok API, and third-party coding platforms and cloud model routers.
Safety and Security
The release includes an entirely new safeguard stack. Grok 4.7 topped LatchBio’s biosafety benchmark at 62.4%. On HackerBench v0.3, a benchmark for malicious cyber tasks, it allowed only 3.3% of risky dual-use prompts through.
xAI has begun offering select cybersecurity partners invite-only access to Grok 4.7’s red-team capabilities for defense research.
Road to Release
The launch caps a drawn-out rollout. Musk posted on September 2 that Grok 4.7 would arrive in 10 days. On September 11 he said it needed a few more days to cook. Grok 4.6 shipped on August 12. Musk had initially suggested Grok 4.7 would follow within four weeks.
Bloomberg reports indicate SpaceX is in talks to acquire data from defunct startups to feed future Grok training. On a recent earnings call, Musk said Grok 5, due later this year, will incorporate the entire corpus of SpaceX data.