Grok 4.7 Explained: xAI's Newest Flagship Model
By ChatGBot · September 30, 2026
xAI is moving fast. Grok 4.1 arrived back in November 2025, and less than a year later the lineup is already on Grok 4.7, released September 21, 2026. It is a bigger base model trained harder on long, multi-hour tasks, and it is aimed at coding and agent work. Here is what is new, how it performs, and where it fits.
Use GPT, Claude, DeepSeek, Qwen and GLM in one app, one subscription.
Try ChatGBotWhat is new in Grok 4.7
xAI says every metric on its own benchmark table improved over Grok 4.6. The clearest jumps are in agentic coding: Terminal-Bench 4.0 climbed to 38.0% from 20.3%, and CursorBench 4.0 rose to 46.3% from 40.4%. The model is described as working longer on difficult tasks and checking its own work more carefully, which is the direction most labs are pushing this year. For how Grok stacks up against other assistants, see Grok vs ChatGPT and Claude vs Grok.
Benchmarks: strong, but not a clean sweep
On xAI's numbers Grok 4.7 leads on some tasks and trails on others. It posts 19.6% on Harvey Legal against a comparison model's 6.7%, and 64.0% on EEBench versus 56.4%. On coding-heavy tests it still trails the strongest rivals, for example CursorBench at 46.3% and Terminal-Bench at 38.0%. In short, it is a serious coding and agent model, but not the outright leader on every benchmark.
Context, pricing, and availability
Grok 4.7 has a 500,000-token context window and a knowledge cutoff of May 2026. In the API it costs $2 per million input tokens and $6 per million output tokens, with cached input at $0.50, and there is a Fast variant at $4 / $12 for lower latency. You can use it in Cursor, Grok Build, and the Grok API, plus third-party platforms like GitHub Copilot, OpenRouter, and Vercel.
How it fits the bigger picture
Grok 4.7 is part of a remarkable few weeks: Claude Opus 5.5, GPT-6 Sol and Luna, GPT-6.1 Sol, and now a new Grok, all within days of each other. The pattern is clear, capable models keep getting cheaper and faster, and the real edge is matching the right model to each task. For the full month in one place, read our September 2026 AI model releases roundup and our take on GPT-6.1 Sol.
Frequently asked questions
Is Grok 4.1 the newest Grok? No. Grok 4.1 launched in November 2025. The current flagship is Grok 4.7, released September 21, 2026.
How much does Grok 4.7 cost? In the API it is $2 per million input tokens and $6 per million output tokens, with cached input at $0.50. A Fast variant costs $4 / $12.
What is Grok 4.7 best at? Long, multi-step coding and agent tasks, where it improves the most over Grok 4.6, though it still trails the top rivals on some coding benchmarks.
What is the context window? 500,000 tokens, with a knowledge cutoff of May 2026.
Which models can I use in ChatGBot? ChatGBot gives you GPT, Claude, DeepSeek, Qwen and GLM under one subscription, so you can switch to the best model for each task.