xAI introduced Grok 4.7 on September 21 as a new model for coding and knowledge work. It is available in Cursor and Grok Build, as well as through the Grok API, third-party coding harnesses, model routers and cloud platforms.
The update is designed to manage longer contexts, review its own output more carefully and work directly with the Grok Bot harness, according to xAI. The company attributes these changes to an expanded base model and an extended reinforcement-learning phase that placed greater weight on difficult, multi-hour assignments.
Pricing starts at $2 per million input tokens and $6 per million output tokens, matching the listed prices for Grok 4.6. xAI also offers a fast variant with twice the output speed at twice the price. The announcement does not specify separate prices or access terms for the named third-party cloud platforms.
In results published by xAI, Grok 4.7 scored 46.3% on CursorBench 4.0, compared with 40.4% for Grok 4.6. On Terminal-Bench 4.0, which covers multi-hour terminal work, the respective scores were 38.0% and 20.3%. Grok 4.7 also reached 71.0% on DeepSWE v1.1 in the table’s high-effort configuration. These are vendor-reported benchmark results and are not presented as independently reproduced.
xAI also says Grok 4.7 was built with an entirely new safeguard stack. The company reports that the model allowed 3.3% of risky dual-use prompts through on HackerBench v0.3 and scored 62.4% on the LatchBio biosafety benchmark. It has also begun giving selected cybersecurity partners invite-only access to the model’s red-team capabilities for defensive research.
Practical context: In practical terms, the update targets more than short code completion: the stated capabilities cover extended agent-style workflows involving terminal use, document preparation and checks on intermediate work. Model selection should weigh performance on the user’s own tasks against output-token costs, because the faster variant also doubles the price.
| Model | Input, $/million tokens | Output, $/million tokens | CursorBench 4.0 |
|---|---|---|---|
| Grok 4.7 xHigh | 2 | 6 | 46.3% |
| Grok 4.6 High | 2 | 6 | 40.4% |
| GPT-5.6 Sol Max | 4 | 20 | 41.7% |
| Fable 5.1 Max | 10 | 50 | 51.8% |
Sources
Event date: 2026-09-21. Primary source date: 2026-09-21.
