What Grok Is
Grok is the AI assistant from xAI (now SpaceXSI), a SpaceX company. It first appeared under the xAI name at the end of 2023, after Elon Musk founded xAI; after SpaceX acquired xAI in 2026, it was folded into SpaceX’s internal AI division, SpaceXAI (read the background here). The models have advanced quickly over the years; the current lineup is summarized below in “The Latest Grok Lineup.”
You can use Grok in two places: the web interface at grok.com, or inside the X (formerly Twitter) app. Grok inside X is embedded directly into the social platform. You can ask it about real-time discussion around a topic, and it scans X posts to answer.
If you mainly want to know how to use Grok in Chinese, what the free plan includes, and which Traditional Chinese prompts to try, see the Grok Chinese Free Guide. This article focuses on what Grok is, its plans, X search, and how it complements ChatGPT.
Grok’s answer style is more direct and less roundabout. Ask a controversial question and ChatGPT may explain both sides without giving a conclusion. Grok is more willing to make a clear judgment.
This trait still exists in 2026. Grok’s “persona” is straightforward and sometimes a little humorous. Whether you like that style is personal preference, but at least it feels different from ChatGPT.
Plans and Pricing
Grok’s paid structure is different from other AI tools. Since October 12, 2026, new subscriptions go through The X Subscription, one plan that covers Grok, Cursor, and X paid benefits, with Grok Bot from Premium up (official announcement). Prices below are from the official pricing page, checked October 13, 2026, in US dollars before tax.
Free (US$0): available on grok.com, X, and the app, with daily access to chat, image generation, and voice. Grok assigns a model automatically; there is no model picker, and usage is limited. It can handle simple Q&A, translation, and short writing.
Plus (US$8/month): the entry-level paid plan, with base usage and an X checkmark. It does not include Grok Bot.
Premium (US$30/month): 4× Plus usage, the latest Grok models, AI video, Grok Bot, and full X Premium+ benefits. Above it are Super (US$100, 15× Plus, priority at peak times) and Ultra (US$200, 50× Plus). Paid plans share one weekly pool for Grok models in Grok and Cursor; Cursor’s third-party models are billed separately, and no message count is published (usage limits). X benefits apply after you connect a personal X account. API context specifications do not describe the actual limits in the chat plan.
Older plans: SuperGrok (including Lite and Heavy) and X Premium are closed to new sign-ups. Existing subscribers keep their plan, price, and features, and can switch in web billing settings once the option appears. Switching cannot be undone; eligible direct-billed subscriptions included in the switch receive a prorated refund for unused time, App Store refunds are Apple’s decision, and Google Play subscribers should check that the old plan stopped renewing (refund details).
The new plans are rolling out region by region, so Taiwan’s actual cost is what checkout shows.
For heavy X users, Premium covers Grok and X Premium+ benefits in one plan, so there is no need for separate subscriptions.
Free vs Paid: How to Decide
| Need | Recommendation | Why |
|---|---|---|
| Try Grok or occasionally check live information | Free plan | Sign in at grok.com for light Q&A and real-time lookup |
| Many daily turns and higher quota | Plus | Solve the quota issue without jumping straight to a higher tier |
| Grok Bot, AI video, or the latest models | Premium | Suits frequent use; long-document context follows the actual interface limits |
| You already use X heavily | Premium | Includes X Premium+ benefits, no separate subscription |
| You often run out of weekly usage | Super / Ultra | 15× / 50× Plus usage |
For detailed free-plan limits and Chinese prompt tips, see Grok Chinese Free Guide.
The Latest Grok Lineup
Current lineup (updated September 29, 2026): The official grok.com plans page lists Grok 4.6 (released August 12, 2026) as the chat model. Free users are assigned a model automatically and cannot choose one; the official pricing page (checked October 13) lists the latest Grok models from Premium up. See What Is Grok 4.6? for a full analysis.
The updated Grok 4.7 (released September 21, 2026) launched first on Cursor, Grok Build, and the API, with API pricing starting at input US$2 / output US$6 per 1M tokens. The official announcement does not mention grok.com or the app. Check the in-app information to see which version the chat interface is using.
Grok has improved substantially compared with its early versions.
Reasoning: The latest Grok has caught up with mainstream tools on logic and complex analysis. In independent evaluations, it has even ranked near the top for completing real-world workplace tasks. Grok and ChatGPT produce similar accuracy on the same math and code-debugging questions. Independent testing found Grok 4.5’s hallucination rate was higher than its predecessor’s; important answers need checking.
Long context (API): Official documentation currently lists a 500,000-token API context window for the current model. A single request reaching 200K tokens is charged at the higher rate for the entire request. Long context is useful for documents, but is no longer unique: flagship APIs are in a similar range, while ChatGPT’s interface context depends on the plan and mode.
Real-time information: Grok can access live posts on X and search the latest web information. Ask what happened to TSMC stock today and it can provide same-day information, unlike tools that only answer from data before their training cutoff.
Code generation: The latest version has improved at coding, and the company promotes the cost efficiency of agentic coding. Python and JavaScript output is usable, though long-running bug-fixing tasks still lag behind Claude Code. It is handy for explaining code you do not understand.
Image and video understanding: Image recognition has improved in recent versions. Grok can also understand video, including reading text in screenshots and analyzing chart data.
Chinese Ability
Grok can handle everyday conversations in Traditional Chinese, but it is weaker at long-form writing and nuanced language tasks than leading AI tools. For polished Chinese output such as articles and reports, use Claude.
Hands-on tests of Chinese reply quality, Traditional Chinese prompt tips, and free-plan limits are covered in the Grok Chinese Free Guide.
X Platform Integration
This is what makes Grok different from other AI tools.
Inside the X app, Grok can do a few things other AI tools cannot:
Analyze real-time discussion. Give it a topic and it scans recent X posts to tell you what people are discussing and what the sentiment looks like. Useful for tracking news and monitoring brand conversation.
Summarize threads. If a long thread is too much to read post by post, send it to Grok for the key points.
Explain memes and image jokes. X has many memes. If you do not understand one, ask Grok and it can explain the background and context.
A practical example is tracking crypto discussion. When a token pumps or dumps, ask Grok “what has X been discussing in the last hour,” and it quickly organizes the main directions of discussion. Much faster than scrolling posts yourself.
Grok as AI Search
Grok is closer to “an AI summary layer over X and real-time web search” than a traditional search engine or rigorous research tool. It fits questions like:
- What are people on X saying about a breaking story?
- What is the recent social sentiment around a brand, coin, or public figure?
- What is the context behind a thread or meme?
It is not ideal for final research conclusions you can cite directly. For source-first verification, pair it with the tools in AI Search Tools Guide, such as Perplexity, ChatGPT Search, or Deep Research workflows.
AI Video Generation
Grok has AI video generation, but the experience is limited. Single short clips can be acceptable. The problem appears when you try to stitch several clips together. Grok cannot deeply integrate video segments, so each one feels generated independently and the transition is stiff. If you expect “give it a script and it makes a complete short video,” it is not there yet.
What is it good for? Occasionally generating a 5-10 second moving visual for social posts. For formal video content, current AI video tools are still not quite enough.
Voice Feature Test
Grok has voice chat inside the app. Operation is simple: hold the microphone button, speak, and Grok replies by voice. Chinese voice input is supported, but voice replies are too stiff. They sound like a prepared script being read aloud, flat in tone and without conversational feeling. Text can be scanned quickly; voice has to be listened to sentence by sentence, so efficiency is worse.
Penchan previously used Grok to practice English slang. It could hear some intonation, but the Q&A had an odd feeling. For example, it might end with something like “try it and I’ll listen,” then repeat the same sentence after every attempt. Still quite rigid.
Compared with ChatGPT voice mode, the gap is obvious. ChatGPT has rhythm and intonation, closer to talking to a person. Grok feels like TTS (text-to-speech).
Typed replies are smooth; that is Grok’s main arena. Voice mode may become more useful if xAI upgrades the voice model later. For now, type directly.
A Practical Set of Grok Uses
Check real-time information: news events, market movement, and discussion direction on X. This is Grok’s strongest scenario and something other AI tools cannot do.
Quick Q&A: similar to using ChatGPT. If the X app is open, it is convenient to ask Grok directly.
Explain code: when you occasionally run into code you do not understand, send it to Grok and ask for an explanation. Any AI can do this; use whichever is convenient. Grok is a good fourth opinion outside ChatGPT/Claude/Gemini.
Not suitable for: long-form writing, formal translation, and software development. Grok is not the first choice for these.
Compared with ChatGPT
| Dimension | Grok | ChatGPT |
|---|---|---|
| Real-time information | Strong, can access X and web | Has search but less immediate than Grok |
| Answer style | Direct, sometimes humorous | More standard and cautious |
| Chinese ability | Usable but weaker | Better than Grok |
| Ecosystem | X platform integration | Plugin store, GPTs |
| Long-form writing | Average | Better than Grok |
| Code | Usable | Better than Grok |
For a more detailed comparison, see Grok vs ChatGPT.
People who use both will find a lot of overlap, but each has one thing the other cannot do: Grok has X integration and real-time information, while ChatGPT has the plugin ecosystem and GPTs.
Practical split: use Grok for live lookup, X sentiment, and blunt second opinions; ChatGPT for brainstorming, general discussion, and plugin tasks; Gemini for Google ecosystem and image generation; Claude for long-form and formal output.
Should You Use Grok?
If you already use X (formerly Twitter), trying Grok is almost zero-cost. The free version works, and in the X app it is one tap away.
Is it worth paying for? That depends on whether you use X’s paid features. The new plans bundle Grok with X benefits (a checkmark on Plus, full X Premium+ benefits from Premium up), so if you already wanted X’s paid features, Grok comes along as a bonus. But if you only want an AI assistant, there are more mature choices on the market.
Grok’s 2026 positioning is clear: it is the AI assistant inside the X ecosystem. If you live in X, Grok is useful. If you do not use X, other AI tools may fit better.
Penchan’s Take
Penchan uses Grok through X Premium+, because X is already part of the daily workflow and there is no separate SuperGrok subscription. In the tool stack, Grok’s role is “real-time information + X discussion sentiment,” a space other AI tools do not cover.
General reply quality is satisfying and the tone feels natural. Voice mode was tested for a few days and then dropped back to typing because it was too stiff. For crypto discussion, Grok is much faster than scrolling X manually.
Long-form writing and software development do not go to Grok. Text output goes to Claude, coding goes to Claude Code. Grok’s role is very clear: fill the X / real-time information slot.
Pricing and features were rechecked on October 13, 2026. xAI and X may adjust plans at any time; confirm with the latest official announcement.