Chat
Grok
The assistant on grok.com, in the Grok apps, and on X. Chat, voice, images, and video. It spends the same weekly pool as Build, Bot, and Grok models in Cursor.
Interactive comparison
Four names, and the other benches.
Grok, Grok Bot, Grok Build, and Cursor are easy to price as four products. For new buyers they are one X Subscription and one weekly usage pool. This page compares that bundle with Claude Code, Codex, GitHub Copilot, Devin Desktop, and Cline.
List prices, before tax
Usage multiples are not the same unit across vendors
Sentiment is a reading of public posts, not a poll
What each name actually is
The chat box, the always-on teammate, the terminal agent, and the editor got bundled on 12 October 2026. Older SuperGrok and Cursor Pro plans still run for people who kept them. Switching is optional, and a completed switch is permanent.
Always-on agents
Grok Bot
Named teammates with their own cloud computer. They use a browser, files, and a terminal, and they keep working after you close the laptop. Included on Premium, Super, and Ultra. Not on Plus.
Terminal agent
Grok Build
The coding agent in your terminal. The command is grok. It also runs headless, or inside other editors over ACP. Subscription usage, or an API key.
Editor
Cursor
The AI code editor. Plus runs Grok models only. Premium adds cloud agents and Bugbot. Claude, GPT, and Gemini in Cursor are billed as extra credits from the first request.
One seat covers all four. Buying “Grok plus Cursor plus Grok Bot” is not three invoices. Bot usage is not a second pool. A heavy Bot week is a lighter week for chat, Build, and Cursor.
Pricing charts
Bars use published list prices. Annual mode uses the vendor’s stated monthly equivalent where one exists. Copilot, Devin, Claude Max, and the Codex figures stay on monthly prices when no annual rate was on the page checked.
Billing
Who pays
Seats
1
Chart
Price to unlock
X products selected
—
Other selected products, added together
—
The X Subscription ladder
Bar height is a log scale of the published multiple, so 4× is visible next to 50×. The dollar figure is the price. The outline follows the stack control below.
Dollars per published multiple
Each vendor’s own base. Do not compare the units.
Copilot credits versus the fee
Official face value. 1 credit = $0.01. Model choice spends them unevenly.
What a stack costs
Pick the jobs and how hard you run them. The three stacks are explicit recipes, not a claim that one is best. Gaps stay visible: a checkmark means the list price includes it, a gap means you still pay extra or only get part of it.
How hard you run it
Stack list price
Plan tables
Every number in the charts comes from these rows. Codex prices are marked because they were cross-checked from pricing trackers on 18–29 Oct 2026, not read off chatgpt.com in this session.
| Product | Plan | Monthly | Annual | What the vendor says you get |
|---|
Capability matrix
Filtered by the product chips in the price section. “Partial” means the tool reaches that job with a limit called out in the cell.
Sentiment, with the context attached
This is a close reading of public posts, mostly Reddit, from 6–29 October 2026. People who are angry or amazed post more than people who are fine. Theme marks are Quiet, Common, or Loud inside that reading. They are not survey scores, and products with a thin reading say so.