Grok 4.7 is here, offering high performance at the same price as the previous generation model.



SpaceXAI announced Grok 4.7 on September 21, 2026. It shows significant improvements at the same price and speed as Grok 4.6, and CEO Elon Musk is promoting it as 'a powerful combination of intelligence, speed, and low cost.'

Introducing Grok 4.7 | SpaceXAI

https://x.ai/news/grok-4-7







The Grok 4.7 is described as 'SpaceXAI's most powerful model ever,' ideal for coding and knowledge-based tasks.

The following graph shows the performance offered by different pricing plans in ' CursorBench 4.0 ,' which puts a load on long-running coding tasks. It demonstrates that Grok 4.7 can achieve performance comparable to competing models such as Claude Opus 5 and GPT-5.6 Sol at a lower cost.



Grok 4.7 is priced at $2 (approximately 315 yen) per million input tokens and $6 (approximately 947 yen) per million output tokens. Compared to Grok 4.6 at the same price point, it performs better in CursorBench 4.0 and other software engineering benchmarks, electrical engineering benchmarks, multi-hour office work, terminal work, and professional tasks. It also outperforms the relatively more expensive 'GPT-5.6 Sol Max' and 'Fable 5.1 Max' in many scores.



SpaceXAI is promoting Grok 4.7 as having superior performance in document and presentation creation. Below are graphs of '

GDPval ' (left), which measures how much specialized tasks AI can handle, and ' AA Briefcase ' (right), which evaluates the long-term and advanced practical capabilities of an AI agent. In both cases, Grok 4.7 surpasses Grok 4.6 and achieves performance on par with other cutting-edge models.



The following is a comparison of creating open-world games using Grok 4.7 and Grok 4.6. The game created with Grok 4.6 in the second video has rough polygon characters and a monotonous background, while the game created with Grok 4.7 in the first video has significantly improved graphics.




Furthermore, Grok 4.7 is built on an entirely new security stack, which SpaceXAI describes as 'the most powerful model we've ever tested in terms of denial and jailbreak resistance.' It demonstrated top-tier security in HackerBench v0.3 , a benchmark for high-risk malicious cyber tasks, allowing only 3.3% of high-risk dual-use prompts to pass through. Moreover, it rarely blocks legitimate security tasks, achieving a balance between robust cyber defenses and a low denial rate for legitimate use.

Grok 4.7 will be available in Cursor and Grok Build starting September 21, 2026. It will also be available through the Grok API, third-party coding harnesses, model routers, and cloud platforms. Pricing starts at $2 (approximately 315 yen) per million input tokens and $6 (approximately 947 yen) per million output tokens, with a high-speed version available that offers double the output speed and double the price.

Related Posts:

in AI, Posted by log1e_dh