GRK
// LLMs · language models

Grok

SpaceXSI’s AI (formerly xAI): a chatbot with a free plan and real-time X search, plus a Grok 4.7 API at $2 per million input tokens.

Pricing
Free plan · Plus $8/mo · Grok 4.7 API at $2 per 1M input tokens and $6 output · prices in dollars
Model
Freemium
Evaluation
External evaluation
Verified
November 16, 2026
◆ Operator verdict

Grok is the chatbot and model family from SpaceX Super Intelligence, formerly xAI, which SpaceX acquired in February 2026. Its current model, Grok 4.7, has a 500,000-token context window and costs $2 per million input tokens and $6 output through the API, and its standout feature is real-time search on X. I can build models and APIs like Grok into business agents and automations, and train the team that will maintain them. What the homepage won’t tell you: the chatbot trains on your conversations unless you turn it off, prices are in dollars and exclude taxes, and the API’s default endpoint doesn’t guarantee which region processes your data.

+ What it’s GOOD for

  • Monitoring X in real time (your brand, competitors, crises, launches): the API ships with the X Search tool, at $5 per 1,000 posts returned
  • Agents and automations that handle long documents: Grok 4.7 supports a 500,000-token context window, and Grok 4.3 and 4.20 support 1 million
  • Teams already using OpenAI-compatible tools: the official docs call Grok 4.7 with the OpenAI SDK by swapping the base URL and the key
  • Trying a frontier model without paying: there’s a Free plan on grok.com, iOS and Android, with limits
  • Anyone already paying for Cursor or X Premium: The X Subscription bundles Grok, Cursor and X into one plan starting at $8 a month
  • Diversifying beyond Anthropic, OpenAI and Google: grok-4.7 costs the same through the API as qwen3.8-max ($2 and $6)

− What it’s NOT for

  • Personal data in the chatbot without checking your settings: on consumer plans, Grok can train on your conversations until you turn off “Improve the Model” or use Private Chat
  • Companies contractually required to process data in the EU: the default endpoint doesn’t guarantee a region, and the only documented regional endpoint is in the US
  • Budgets that need a fixed price in euros: SpaceXSI publishes its prices in dollars, excluding taxes
  • Brands or industries with a reputational veto on SpaceX or X: that’s not a technical call, and it’s worth making before you integrate
  • Prompts over 200,000 tokens without running the numbers: past that threshold, the whole request is billed at double the rate

◇ Where it fits across StratoAI lines

AI strategist at night in front of two monitors, one showing a chat conversation and the other a terminal making API calls, next to a notebook with a diagram of one model splitting into a chat app and an API

Grok is the AI from SpaceX Super Intelligence (SpaceXSI), formerly xAI: a chatbot you can use for free on grok.com, in the iOS and Android apps or inside X, and a family of models you can integrate through an API. The current model is Grok 4.7, released on September 21, 2026, with a 500,000-token context window, and through the API it costs $2 per million input tokens and $6 output. I can build models and APIs like Grok into business agents and automations, and teach your team how to maintain them. Here’s who makes it today, what each piece costs, what happens to your data and when I wouldn’t recommend it.

Who makes Grok today

If you’re looking for “xAI,” it goes by a different name now. On February 2, 2026, the company’s own site announced it in one line: “SpaceX announced today that it has acquired xAI.” Today, the official Grok pages carry the footer “© 2026 SpaceX Super Intelligence LLC,” and the brand calls itself SpaceXSI (the developer docs still live at docs.x.ai and the API at api.x.ai).

One point most articles skip: Grok lives inside X, but it isn’t the same company. The legal FAQ puts it this way: “SpaceXSI is a separate company from X Corp. (“X”, previously Twitter).” And for Europe, the privacy policy addendum states that “SpaceX Super Intelligence LLC, which is based in the USA, is the controller,” meaning the US company is the data controller.

What about Elon Musk? None of the official pages I read for this review names anyone at the helm of SpaceXSI, so I won’t tell you who runs it today.

What Grok is (and which name means what)

The same name covers a chatbot, several models and products that aren’t chatbots at all. Here’s the map:

NameWhat it isWhere
Grok (chatbot)The assistant: it chats, searches the web and X in real time, reasons, analyzes files and generates images, video and codegrok.com, the iOS and Android apps, and inside X
Grok 4.7 and familyThe language models you integrate through the APIdocs.x.ai
Grok ImagineImage and video generation, in the chatbot and through the APIInside Grok
Grok VoiceVoice conversations and voice APIsInside Grok and through the API
Grok BuildCoding agent for the terminal, the web and mobileOutside this review
Grok BotAgents with their own cloud computer that work in your appsGrok Bot review
The X SubscriptionThe single plan that bundles Grok, Cursor and X, since October 12, 2026This review, below
SuperGrokGrok’s previous paid plan; existing subscribers keep itThis review, below

According to its official page, the chatbot searches the web and X live, has a mode where several agents work in parallel, remembers across conversations, analyzes PDFs and images, does voice, and supports 30+ languages. I haven’t measured its quality in Spanish with a test of my own.

Grok 4.7 and the current model family

Grok 4.7 shipped on September 21, 2026. The docs describe it as “SpaceXSI’s frontier model built for coding, agentic tasks, and knowledge work.” According to the official announcement, it uses a larger base model than Grok 4.6, is served “at the same price and speed as Grok 4.6,” and was trained to natively understand the Grok Bot harness.

ModelReleasedContextAPI (input / output per 1M tokens)Notes
grok-4.7Sep 21, 2026500,000$2 / $6Text and image in, text out. Adjustable reasoning (low, medium, high by default, xhigh). Knowledge cutoff May 2026. Default model in Grok Build
grok-4.6Aug 12, 2026500,000$2 / $6Also available on Amazon Bedrock, Microsoft Foundry, GitHub Copilot and Gemini Enterprise Agent Platform
grok-4.5Jul 16, 2026500,000$2 / $6Reached grok.com, X, iOS and Android on July 22
grok-4.3 · grok-4.20 (reasoning, non-reasoning and multi-agent)4.20: March 20261,000,000$1.25 / $2.50Cheaper, with more context
grok-build-0.1May 25, 2026 (beta)256,000$1 / $2The model behind the Grok Build coding agent

Where Grok 4.7 runs: on the SpaceXSI API, on a US regional endpoint (10% more expensive), in Grok Build, in Cursor (“available on all plans”) and through the OpenRouter, Vercel and Cloudflare gateways. There’s also Grok 4.7 Fast, the same model on faster infrastructure at twice the price, but only in Cursor and Grok Build, not on the public API.

A word of caution on benchmarks. The 4.7 announcement pitches it as “Twice as fast, at half the price of comparable models” and publishes its own results: 46.3% on CursorBench 4.0 (at xhigh effort) and 71.0% on DeepSWE (at high effort). In that same vendor table, Fable 5.1 comes out ahead on CursorBench (51.8%) and GPT-5.6 Sol on DeepSWE (72.7%). These are self-reported numbers, and I don’t have an independent comparison I can cite.

Which model does the grok.com chatbot use today? The last official announcement that says so is from July 22, 2026: “Grok 4.5, our most intelligent model yet, is now on grok.com, X, iOS, and Android.” I haven’t found a statement saying Grok 4.7 is now the chatbot’s default model, so I’m not assuming it.

Is there a Grok 5?

No Grok 5 has been announced. It doesn’t appear on any official page I reviewed (the x.ai news feed, the docs, the release notes and the About page). The latest model is Grok 4.7, from September 2026. What’s out there about Grok 5 is rumor or speculation, not a product you can use or a date that’s been announced.

How to use Grok and what it costs

To use it, go to grok.com, download the iOS or Android app, or open Grok inside X. The official site sums it up as “Free to try on the web and in the apps.” You sign in with your X account or your email (the official help docs also mention Google and Apple), and your conversations sync across devices.

To pay for it, since October 12, 2026 new customers sign up for The X Subscription, a single plan that bundles Grok, Cursor and X perks. Prices from x.ai/pricing and its documentation, checked on November 16, 2026, in dollars:

PlanMonthlyAnnualIncluded usageWhat it adds
Free$0—LimitedChat and agent requests with limits; basic X features
Plus$8$80BaseGrok Chat, Voice, Imagine and Build, Grok API access and Cursor; the X Premium checkmark and fewer ads. No Grok Bot
Premium$30$3004x PlusGrok Bot, the latest Grok models, faster responses, images and videos, cloud agents and automations
Super$100$1,00015x PlusMuch more usage across chat, Imagine, voice and Build, 1080p 15-second videos and priority access at peak times
Ultra$200$2,00050x PlusEverything above, with the most usage headroom
TeamsFrom $30 per user—Per seatPremium, Super or Ultra seats with centralized billing and admin. Plus isn’t available for teams
EnterpriseCustom—CustomThrough sales: custom rate limits, SSO, data residency and volume pricing
Open notebook on a wooden desk at night with a hand-drawn sketch of five tiers rising like a staircase, next to a mechanical calculator, glasses, a pen and a cup of coffee, with a monitor showing a bar chart
Five plans in a staircase: what mostly changes between them is how much weekly usage you get.

What to know before you pay:

  • One weekly usage pool. Grok and Cursor usage draw from the same pool, which resets every week. When you run out, you buy additional credits or wait for the reset. Third-party models in Cursor (Claude or GPT, for example) are billed separately from the first use.
  • The API is billed separately. Paid plans include API access, but as x.ai/pricing puts it, “API usage is billed separately at standard model rates and doesn’t draw from your plan’s usage pool.” API prices are below.
  • Grok Bot starts at Premium. Free and Plus don’t include it, and its usage comes out of the same weekly pool.
  • The usage multipliers, page by page. x.ai/pricing lists the included usage of Premium, Super and Ultra as 4, 15 and 50 times Plus. The Grok Bot page (x.ai/bot) states it relative to Premium: Super gets “3.5x Premium usage” and Ultra “12x Premium usage.” Both figures are official, but they don’t quite add up (15 divided by 4 is 3.75, and 50 divided by 4 is 12.5), so treat them as approximate.
  • If you already pay for SuperGrok. Existing SuperGrok, Cursor and X Premium plans keep working, and switching is optional. But according to the docs, switching to The X Subscription is permanent: it can’t be undone. Grok’s own page still invites you to “Upgrade to SuperGrok,” even though new customers now start on The X Subscription. SuperGrok’s current price doesn’t appear on the official pages I reviewed.
  • Direct sales only. x.ai warns that subscriptions are only sold through X Subscription Corp. and that accounts bought from resellers may be suspended.

◇ Price in euros and VAT: not published. x.ai/pricing lists prices in dollars and notes: “Prices shown do not include applicable taxes or fees (e.g. VAT).” I couldn’t confirm the final price in euros or the countries where each plan is sold. Before you budget, check the total at checkout.

Grok API pricing

Prices from the official API pricing table, checked on November 16, 2026. The page itself notes: “All prices are in USD.” These are dollars per million tokens:

ModelContextInputCached inputOutput
grok-4.7 (prompt under 200,000 tokens)500,000$2.00$0.50$6.00
grok-4.7 (prompt of 200,000 tokens or more)500,000$4.00$1.00$12.00
grok-4.6 · grok-4.5 (under 200,000)500,000$2.00$0.50 · $0.30$6.00
grok-4.3 · grok-4.20 (under 200,000)1,000,000$1.25$0.20$2.50
grok-build-0.1 (under 200,000)256,000$1.00$0.20$2.00
  • The long-context rate applies to the whole request. If the prompt reaches the 200,000-token threshold, the docs say the request is billed “at the higher rate for all tokens in the request,” not just the part over the threshold.
  • Discounts and surcharges. The Batch API takes 20% off the Grok 4.x models in the table (not grok-build-0.1); priority processing costs twice as much; the US regional endpoint adds 10% (grok-4.7 comes to $2.20 and $6.60).
  • Tools. Web search costs $5 per 1,000 calls; X Search, $5 per 1,000 posts and $10 per 1,000 profiles returned; code execution, $5 per 1,000 calls. Without these tools, the model knows nothing after its training data.
  • grok-4.7 rate limits. 150 requests per second and 50 million tokens per minute.
  • Works with what you already use. The official Grok 4.7 guide includes examples with the OpenAI SDK pointed at https://api.x.ai/v1, with Vercel’s AI SDK and with its own Python SDK.

Grok Imagine: images and video

Grok Imagine is the part that generates images and video from text or a reference photo. In the chatbot, Plus includes it, Premium highlights image and video creation, and 1080p 15-second videos show up on Super. Through the API, grok-imagine-image-2.0 costs $0.04 per image and grok-imagine-video-1.5 costs $0.08 per second of video (the “lite” version, $0.02).

Privacy: does Grok train on your data?

There’s a big difference here between the chatbot and the API, and it’s worth reading before you paste in anything from a client.

In the chatbot (consumer plans), yes, unless you turn it off. The official FAQ is blunt about it: “We may use your content and interactions with Grok (e.g., prompts, searches, and other materials you submit) along with Grok’s responses to train our models.” It adds: “You control whether your data is used for training Grok.” Here’s how to turn it off:

  • On grok.com: Settings → Data → “Improve the Model.”
  • In the app: Settings → Data Controls → “Improve the model.”
  • Private Chat (the ghost icon, top right): according to the FAQ, those conversations aren’t used for training and are deleted from SpaceXSI’s systems within 30 days.
  • Without signing in, watch out: “in some regions (excluding the EU/UK)… you won’t have the option to opt out of model training.”

The same FAQ notes that a limited number of authorized personnel may review conversations for specific purposes, and asks: “please do not share any personal or sensitive information in your questions to Grok.” Turning off training applies to new conversations. And the chatbot’s Build mode doesn’t support Private Chat.

Hand holding a phone with a settings screen and a switched-off toggle, in front of a laptop with an open chat and a notebook with a hand-drawn padlock, on a desk at night
The first thing worth checking on a Grok account: the toggle for training on your data.

In the API, no, unless you allow it. The security FAQ says: “SpaceXSI never trains on your API inputs or outputs without your explicit permission.” By default, requests and responses are stored encrypted for 30 days to audit abuse and then deleted; there’s a Zero Data Retention (ZDR) option, though it disables features such as the Batch API and files. For business customers, the legal FAQ also says their content isn’t used to improve the models.

Where API data is processed. The same FAQ admits the default endpoint “does not guarantee a processing region.” To stay in the US, there’s the us.api.x.ai regional endpoint (grok-4.7 and grok-4.6 only, 10% more expensive). I haven’t found an EU regional endpoint; the only verified fact for Europe is that Grok 4.5 reached the API console for EU users in July 2026. Asked in its own FAQ about GDPR and SOC 2, the official answer only states “We are SOC 2 Type 2 compliant.”

What you generate is yours. The FAQ says “You own Grok’s Outputs” and allows commercial use, though it asks you to attribute generated work to Grok.

Grok isn’t Grok Bot

This review covers the Grok chatbot and its models. If what you’re after is the agent that works for you on a cloud computer (email, calendar, Slack, reports), that’s Grok Bot, a separate product that came out of beta with version 1.0 on November 2, 2026. You can’t buy it on its own: it’s included in the Premium ($30 a month), Super and Ultra plans of The X Subscription, and its usage comes out of the same weekly pool as the chatbot. Free and Plus don’t include it.

The two are closely linked: Grok 4.7 was trained to understand the Grok Bot harness. But the chatbot talks with you, while Grok Bot gets work done inside your apps. Its review covers what it does, what it costs, its alternatives and my experience implementing it.

Grok alternatives

Entry prices from each vendor’s official pages, checked on November 16, 2026. For the API, dollars per million tokens (input / output); for the chatbot, the entry-level paid plan.

AlternativeChatbot (entry plan)Reference APINote
Grok (baseline)Free · Plus $8/mogrok-4.7: $2 / $6Excludes taxes
Claude (Anthropic)Free · Pro $20/mo ($17 billed annually)Sonnet 5.5: $2 / $10 · Haiku 5.5: $0.10 / $0.50 · Opus 5.5: $4 / $20Excludes taxes. Claude review
OpenAINot confirmed today (I couldn’t read the ChatGPT pricing page)gpt-6.1-sol: $2 / $10 · gpt-6-luna: $0.10 / $0.50OpenAI review
Gemini (Google)Free · AI Plus €4.99/mo · AI Pro €21.99/moGemini 4 Hydrogen: $1.50 / $7.50Chatbot prices in euros, from Google’s Spanish page. Gemini review
MistralVibe (formerly Le Chat) Free · Pro $14.99/mo excluding taxesMistral Large 4: $0.70 / $2.10 · Small 4: $0.15 / $0.60The European option. Mistral review
Qwen (Alibaba)Qwen Studio free · Token Plan from $6/moqwen3.8-max: $2 / $6 · qwen3.8-flash: $0.15 / $0.47Singapore list prices. Qwen review
DeepSeek—deepseek-flash: $0.15 / $0.60 off-peak ($0.30 / $1.20 at peak)DeepSeek review

Reading the prices. Grok 4.7 costs the same as qwen3.8-max ($2 and $6) and matches Claude Sonnet 5.5 and gpt-6.1-sol on input, but with cheaper output ($6 vs. $10). It’s pricier than Mistral Large 4 and the fast models from Qwen and DeepSeek; against Gemini 4 Hydrogen ($1.50 / $7.50), it charges more for input and less for output; and it’s far cheaper than the top models from Anthropic and OpenAI.

Grok or ChatGPT? It’s a comparison with real search demand (“grok vs chatgpt”: 210 searches a month in Spain and 8,100 in the US, per DataForSEO as of November 16, 2026), and I’m not going to crown a winner on quality: I don’t have an independent test to cite, and each vendor’s benchmarks are published by the vendor. What I can tell you is where they differ. Grok comes with real-time X search built in and an $8 entry plan that includes API access; ChatGPT has its own ecosystem, which I cover in the OpenAI review. The honest way to decide is to try both on your real tasks.

My operator’s verdict

Grok is a serious, well-priced option today: Grok 4.7 gives you a 500,000-token context window at $2 and $6 per million, the API plugs into OpenAI-compatible tools, and it ships with real-time X search built in. If your use case depends on knowing what people are saying on X right now, it’s the first model I’d look at.

What I won’t hide: the chatbot trains on your conversations by default, prices are in dollars and exclude taxes, the API doesn’t guarantee a region except in the US, and the name shuffle (xAI, SpaceXSI, SuperGrok, The X Subscription) would confuse anyone. My advice: try the chatbot with training turned off, measure a real workflow on the grok-4.7 API with prompts under 200,000 tokens and, if there’s customer personal data involved, get the guarantees in writing before you send anything.

I’ll integrate it and teach you how to use it

I can build models and APIs like Grok into business agents and automations, and train the teams who’ll maintain them. The job would be picking the model that fits your data and your budget, costing it against your real volume and leaving it running without you depending on me. And if what you need is an agent that works inside your apps, look at Grok Bot: its review explains how I implement it.

Grok FAQ

What is Grok?

It’s the AI assistant from SpaceX Super Intelligence (SpaceXSI), formerly xAI. You use it as a chatbot on grok.com, in the iOS and Android apps or inside X, and its models can be integrated through an API. The current model is Grok 4.7, from September 2026.

Who came up with Grok AI?

Grok was already xAI’s assistant before SpaceX announced it had acquired the company on February 2, 2026 (“SpaceX announced today that it has acquired xAI”). Today it’s published by SpaceX Super Intelligence LLC, which its legal FAQ describes as a separate company from X Corp.

How do I use Grok?

Go to grok.com, download the iOS or Android app, or open Grok inside X. Sign in with your X account or your email (the official help docs also mention Google and Apple) and start typing. You can switch modes to search, reason, talk by voice or create images.

Is Grok free?

There’s a Free plan with usage limits on chat and agent requests. For more usage, The X Subscription starts at $8 a month (Plus). If you’re looking for a free AI chatbot that works in Spanish, Claude, Gemini, Mistral and Qwen also have free chat; Grok claims 30+ languages, but I haven’t measured its quality in Spanish.

How much does Grok cost?

On The X Subscription: Free $0, Plus $8, Premium $30, Super $100 and Ultra $200 a month, in dollars and excluding taxes. Through the API, Grok 4.7 costs $2 per million input tokens and $6 output (double if the prompt reaches 200,000 tokens). The price in euros with VAT isn’t published.

What is SuperGrok?

It’s the paid plan Grok had before The X Subscription. Existing subscribers keep it and can switch if they want, knowing the switch is permanent. New customers start on The X Subscription; SuperGrok’s current price doesn’t appear on the official pages I reviewed.

How much does the Grok API cost?

grok-4.7 costs $2 per million input tokens, $0.50 cached and $6 output; if the prompt reaches 200,000 tokens, the whole request moves to $4, $1 and $12. Grok 4.3 and 4.20 cost $1.25 and $2.50. The Batch API takes 20% off the Grok 4.x models, and X Search is billed separately: $5 per 1,000 posts.

Does Grok train on my data?

In the chatbot, yes by default: you can turn it off under Settings → Data → “Improve the Model” (Data Controls in the app) or use Private Chat. In the API, no: “SpaceXSI never trains on your API inputs or outputs without your explicit permission.” For business customers, their content isn’t used to improve the models either.

What’s the difference between Grok and Grok Bot?

Grok is the chatbot and the model family; Grok Bot is a team of agents with a cloud computer that works inside your apps. Grok Bot is included from the Premium plan ($30 a month) of The X Subscription.

Is there a Grok 5?

No Grok 5 has been announced: no official source mentions it. The latest model is Grok 4.7, released on September 21, 2026.

What is Grok Imagine?

The part of Grok that generates images and video from text or a photo. In the chatbot it’s included from the Plus plan; through the API, an image costs $0.04 with Image 2.0 and video costs $0.08 per second with Video 1.5.

How accurate is Grok AI?

There’s no neutral accuracy figure, only benchmarks published by the vendor itself. The official FAQ warns that Grok “may at times include in its responses misleading or factually incorrect information” and asks you to verify its answers before using them. With any model, double-check what matters.

What does Grok mean in programming?

In searches, “grok” gets mixed up with other things: it’s also an English verb, and ngrok is a developer service with no connection to SpaceXSI.

Can you help me integrate Grok at my company?

Yes. I can build models and APIs like Grok into agents and automations, and train the team who’ll use them. You can see how I approach AI agents or reach out directly.


Official sources checked on November 16, 2026: xAI joins SpaceX · Grok · The X Subscription pricing and its documentation · What is The X Subscription · Usage limits · Grok 4.7 announcement · Grok 4.7 guide · Models · API pricing · Release notes · x.ai news · Legal FAQ · Privacy policy and European addendum · API security FAQ. Alternatives pricing: Claude and its API · OpenAI · Gemini and its API · Mistral and its API · QwenCloud · DeepSeek.

[ Why StratoAI ]

StratoAI is a cross-industry AI implementation firm. An active portfolio of clients in real estate, specialist retail franchises, restaurant groups (AI work, not hospitality consulting), architecture, B2B distribution, private clinics and D2C ecommerce. A real production stack: Claude Code, n8n and agents running 24/7 — no demos, no slide decks.

A real case, anonymized under NDA: Port concession — diagnosis, business plan, brand identity, renders and website delivered in 15 days and submitted five days before the concession deadline. Executed with Focus Sprint™, the flagship methodology.

Want to bring Grok into your operations?

If this tool fits your case, I’ll implement it inside one of StratoAI’s four lines: Signals, Automations, AI Agents or MicroSaaS. And if it doesn’t fit, I’ll tell you straight. A 15-minute Contact Call, no strings attached.

// Frontier stack

Models I evaluate · providers I run in production

Frontier + open-weight stack · continuous benchmarking · no vendor bias

Want to talk? I reply personally John Guerrero · StratoAI founder