Grok 4.5 Live in Cursor: How to Save 50% on API Tokens

On July 8, 2026, Elon Musk’s artificial intelligence startup xAI officially unveiled its next-generation frontier model: Grok 4.5. This new release represents a massive leap forward in programming capability, mathematical reasoning, and multi-file agentic workflows.

Almost immediately, the development team at Cursor—the world’s leading AI-powered code editor—integrated Grok 4.5 into their platform.

This guide covers how to get started with Grok 4.5 directly inside Cursor, and explains how you can leverage Prompt Caching to save up to 50% on your input token costs when calling the xAI API.

1. Why Use Grok 4.5 Inside Cursor AI?

Grok 4.5 is an Opus-class model designed to serve as an exceptionally powerful coding assistant. Its key strengths include:

Superior Agentic Workflows: Its ability to navigate complex file directory indexing, execute test suites, and orchestrate edits across multiple files in Cursor’s Composer (Cmd+I) is significantly improved.
Low Latency Coding: Compared to comparable models like Claude 3.5 Sonnet or GPT-4o, Grok 4.5 returns its first line of code noticeably faster.
Double Usage Limit (Launch Week Promo): To celebrate the launch, Cursor is doubling the fast request limits for the grok-4.5 model for all Pro and Business subscribers during its first week of availability!

2. Setting Up Grok 4.5 in Your Cursor Editor

To begin coding with Grok 4.5, follow these simple steps to enable it in your editor:

Step 1: Open Cursor and access the settings panel by pressing Cmd + , (on macOS) or Ctrl + , (on Windows).
Step 2: Click on Models in the left sidebar.
Step 3: Under the model list, toggle the switch to enable grok-4.5.
Step 4: Return to your code environment, open the chat sidebar or Composer, and select Grok 4.5 from the model dropdown menu.

3. How to Save 50% on Token Costs via Prompt Caching

If you are using the “Bring Your Own Key” (BYOK) model in Cursor with your personal xAI API key, you can take advantage of xAI’s native Prompt Caching to cut your input token costs in half.

How Prompt Caching Works

When writing code, Cursor continuously sends a large amount of context, including opened files, file path trees, and chat history.
With Prompt Caching, if the prefix of your prompt (like system instructions, developer rules, or previous turns in the chat history) matches a recent query, the xAI servers read it from the cache.
Tokens served from the cache (Cached Input Tokens) are priced at just $1.00 per million tokens, a 50% discount compared to the standard rate of $2.00 per million tokens.

Best Practices to Maximize Cache Hits in Cursor:

Use a Unified System Prompt: Set up a consistent .cursorrules file at the root of your project. This ensures that your system rules are reused and cached across all requests.
Keep Long Chat Threads: Continue asking follow-up questions in the same chat thread rather than starting new ones. The history of that thread will be cached, giving you a 50% discount on the input history of subsequent questions.

4. Get a 50% Cursor Pro Discount on GiftFair

If you want to upgrade to Cursor’s official Pro subscription to enjoy unlimited private usage and double launch-week limits for Grok 4.5, make sure to visit the Cursor AI section on GiftFair.

Using a community-shared referral link will get you 50% off your first month of Cursor Pro (dropping the price to $10 instead of $20). You can also share your personal referral link on GiftFair. When others use it, they get 50% off, and you receive $25 in API usage credits directly in your Cursor account!

Happy vibe coding, and enjoy the speed and reasoning of Grok 4.5!