xAI officially announced the release of Grok 4.7, its latest coding and knowledge-work model, featuring stronger capabilities. The model runs longer on difficult tasks, checks its own work more carefully, and comes equipped with the best safety guardrails to date. Grok 4.7's pricing and speed are the same as Grok 4.6, making it highly competitive.
Grok 4.7 model improvements
Grok 4.7 uses a larger and more advanced base model than Grok 4.6. The model underwent longer reinforcement-learning training focused on complex tasks requiring hours to complete, and performs better at verifying its own work and managing long contexts. Grok 4.7 was also trained to natively understand Grok Bot, enabling stronger performance on conversational tasks and general knowledge work.
「Grok 4.7 is our most capable model for coding and knowledge work.」
xAI
Safety and cybersecurity
Grok 4.7 is equipped with an all-new safety guardrail system, the strongest model we have tested, capable of effectively resisting denial-of-service attacks and jailbreaks. The model excels in dual-use domains such as cybersecurity and biological work, demonstrating high utility on benign tasks and safe refusal on dangerous ones, scoring 62.4% on LatchBio's biosafety benchmark.
「Grok 4.7 performed best in HackerBench v0.3 safety testing, allowing only 3.3% of dangerous dual-use prompts through.」
xAI
Pricing and availability
Grok 4.7 is now available on the Cursor and Grok Build platforms, and accessible via the Grok API, third-party coding tools, and cloud platforms. The model's pricing starts at $2 (USD) per million input tokens and $6 (USD) per million output tokens. A fast version is also offered, with twice the output speed at twice the price.
Source: xAI official announcement