Grok 4 Pricing: What the Grok 4 Tier Costs in 2026

How much does Grok 4 cost?

Grok 4 pricing works differently from what most people expect: Grok 4 has no price of its own. xAI has never sold a standalone Grok 4 subscription; frontier model access is bundled into its tiers, which as of July 2026 are Free at $0 a month, SuperGrok at $30 a month, Grok Business at $30 a month, and Enterprise on a custom quote. Developers pay per token instead, at $2 per million input tokens and $6 per million output tokens on the current flagship.

So the accurate answer to "how much does Grok 4 cost" is $30 a month, because SuperGrok is the tier that has carried the flagship Grok model since the Grok 4 generation launched. That has been true through several model revisions. What changes at each release is which model version sits inside the $30 plan, not the price of the plan.

One correction that matters before you budget: Grok 4 itself is no longer the current model. xAI's flagship in July 2026 is Grok 4.5, and every tier in xAI's public comparison table, including the free one, lists Grok 4.5 access. If you came here to price a Grok 4 seat specifically, what you will actually buy today is a Grok 4.5 seat at the same $30.

Is Grok 4 still the current model?

No. Grok 4 shipped in mid-2025 as xAI's flagship and has been succeeded twice over since. The public lineage runs Grok 4, then a 4.20 revision in early 2026, then 4.3, and then Grok 4.5, which xAI announced in July 2026 and describes as its flagship for code and everything else. xAI's model documentation, last updated 9 July 2026, lists Grok 4.5 as the recommendation for both chat and code and does not list Grok 4 as a separate purchasable model.

This matters for a pricing decision in a specific way. Search volume for "Grok 4 pricing" is real and steady, but the thing people are searching for has quietly been replaced by its successor at the same price point. You are not choosing between paying for Grok 4 and paying for Grok 4.5. You are choosing whether to pay $30 a month for the current frontier Grok, which today happens to be 4.5.

The practical read: ignore the version number when you budget and watch it when you build. Subscription buyers get whatever the current model is, automatically, for the same monthly rate. API consumers have a real choice, because xAI publishes model aliases where the bare model name points at the latest stable version and a dated name pins you to a specific release for workflows that need consistency.

• Grok 4 launched mid-2025 and is no longer the model xAI recommends or lists as flagship.
• Grok 4.5 is the current flagship, with a 500,000-token context window and configurable reasoning.
• Grok 4.5's training knowledge cut-off is 1 February 2026, so it needs search tools enabled for anything more recent.
• Every published tier, including Free, lists Grok 4.5 access in xAI's comparison table.
• API users can pin a dated model version; subscription users always get the current one.

Which plans include the Grok 4 line?

All of them, which is the surprising part. xAI's plan comparison table shows a checkmark for Grok 4.5 on every column: Free, SuperGrok Lite, SuperGrok, SuperGrok Heavy, Business, and Enterprise. The company is not gating the frontier model behind the paywall the way several competitors do. It is gating how much of it you can use.

That changes how you should read the price ladder. The question is not "which tier gets me the good model," because they all do. The question is how many queries you need per day, whether you need Expert-level live search rather than the limited free version, whether you want image and video generation, and whether you need admin controls. Those four things are what the money buys.

It also means the free tier is a legitimate evaluation of the model itself rather than a demo of a weaker one. If Grok 4.5's answers are not good enough for your work on the free plan, paying $30 will not fix that, because it is the same model. Paying fixes running out of room, not running out of quality.

What SuperGrok's $30 a month buys

SuperGrok is the tier most people mean when they ask about Grok 4 pricing, and it costs $30 a month. xAI lists six inclusions on the card: the Grok 4.5 model, connectors, higher rate limits across all features, Expert search, SOC 2 Type I and II compliance, and image and video generation.

The headline benefit is headroom. Free users hit caps; SuperGrok users mostly do not, which is the difference between an assistant you reach for and one you ration. The second benefit is search depth. xAI marks free search as limited and paid search as Expert, and Grok's search reaches X in real time as well as the open web, which is a capability no other major assistant matches on the same terms.

Set against the market, $30 is $10 above the standard consumer AI subscription price. That premium is not buying a better model than the free tier offers. It is buying volume, generation features, and live data. Decide whether those three things are worth $360 a year to you specifically, because that is the real comparison, not Grok versus ChatGPT in the abstract.

SuperGrok Heavy and the multi-agent tier

SuperGrok Heavy is the top consumer tier in xAI's comparison table, sitting above standard SuperGrok. Its price is not printed on xAI's pricing page as of July 2026, and this article does not repeat the figures circulating on third-party sites because they conflict with each other and several appear to describe expired promotional rates.

What Heavy provides is architectural rather than a bigger model. xAI describes a multi-agent mode where several agents work in parallel on sub-problems, each showing auditable reasoning, with results merged into one cited answer. Running many agents on one question costs several times the compute of a single-pass response, and that cost is what the tier prices in.

The buying test is narrow. Heavy earns its price for research-grade questions where a single answer is not trustworthy enough and you would otherwise run the query several ways yourself. It does not earn its price for drafting, summarizing, or lookup work, which the $30 tier handles with the same model. If you cannot name three questions from last month that would have benefited, you do not need it.

• Positioned above SuperGrok as the top consumer tier in xAI's plan comparison.
• Runs multiple agents in parallel, each showing its reasoning, merged into one cited answer.
• Price is not published on xAI's pricing page; read it at checkout rather than trusting secondary sources.
• Worth it only for genuinely hard research questions, not for everyday drafting or lookup.
• Same underlying flagship model as the cheaper tiers, applied with more compute per question.

Grok 4 API pricing

If you are building on Grok rather than chatting with it, the subscription tiers are irrelevant. xAI's API is usage-based with no minimums, and Grok 4.5 is billed at $2.00 per million input tokens and $6.00 per million output tokens with a 500,000-token context window. Those figures come from xAI's own model documentation, last updated 9 July 2026.

Voice and media sit on separate meters. The Voice API is $3.00 per hour for a live agent, $15.00 per million characters for text-to-speech, $0.10 per hour for batch speech-to-text and $0.20 per hour streaming. The Imagine API charges $0.02 per image and $0.05 per second of video.

The number that surprises people is how small per-answer inference cost is. A support-style reply that reads a couple of thousand tokens of retrieved context and writes a few hundred tokens back costs well under a cent at these rates. Anyone modelling the economics of an AI product should internalize that early, because it means the model bill is almost never the reason a deployment is expensive.

Grok 4 pricing vs the general Grok tiers

People searching for Grok 4 pricing usually assume there is a premium model tier separate from the standard Grok subscription, the way some vendors charge extra for their reasoning models. xAI does not work that way. There is one consumer ladder, the frontier model is available on every rung including the free one, and the rungs are defined by usage and features.

The closest thing to a "Grok 4 tier" is SuperGrok at $30, because that is where the frontier model became practical to use at volume when the Grok 4 generation shipped, and it remains the tier where a serious individual user lands. Everything above it is either multi-agent depth or organizational controls.

For comparison against the wider market: ChatGPT Plus is $20 a month, ChatGPT Go is $8, ChatGPT Pro starts at $100, and ChatGPT Business is $20 per user per month billed annually or $25 billed monthly. Grok's $30 sits between Plus and Pro, and its free tier is more generous than most. There is no clean winner on price; there is a clean difference in what the premium buys.

That difference is worth naming precisely, because it decides the purchase. Grok's paid tiers are selling live data: Expert-depth search across the open web and X, in real time, with citations. Competing subscriptions are mostly selling model access and connector ecosystems. If your work turns on what is happening right now, the $10 premium over Plus is doing real work every day. If it turns on reasoning over documents you already have, you are paying more for a capability you will not open.

What Grok 4 costs for teams and businesses

Grok Business is listed at $30 a month on the Team view of xAI's pricing page and covers all Grok models including 4.5. It adds Grok Build access, team seat management, consolidated billing, role-based access control, domain verification, user analytics, SOC 2 Type I and II compliance, and priority support.

The card does not state whether $30 is per seat or per workspace. Given that the plan is built around seat management and consolidated billing, per-seat is the reasonable assumption, but do not budget on an assumption. Confirm the billing unit with xAI before you size a deployment, because that single detail changes a forty-person rollout by an order of magnitude.

Enterprise is quote-only. What it adds is governance: custom SSO, Directory Sync via SCIM, custom role-based access control, advanced audit controls, customer-managed encryption keys, a dedicated data plane, data residency, dedicated onboarding, and volume discounts. xAI routes those conversations through its sales team. Expect the same dynamics as any enterprise software negotiation, where seat count, term length, and timing all move the number more than the list price does.

• Business adds seats, RBAC, domain verification, consolidated billing, and user analytics for $30 a month as listed.
• Confirm whether that $30 is per seat before sizing a rollout; the card does not say.
• Enterprise is negotiated and adds SSO, SCIM, audit controls, customer-managed keys, and a dedicated data plane.
• Both tiers list a no-training commitment on your data alongside SOC 2 Type I and II compliance.
• Neither tier is a customer-facing product; both licence employees to use Grok internally.

What the Grok price list does not cover

A published price is not the same as a total cost, and there are four gaps between xAI's pricing page and what you will actually spend. None of them are hidden fees in the dishonest sense; they are just things a two-card pricing page cannot express, and they catch people out consistently.

The first is that subscriptions and API credits are separate purchases. Paying $30 a month for SuperGrok gives you nothing in the API console, and API spend gives you nothing in the Grok app. Teams routinely budget one and get surprised by the other. The second is platform pricing: buying through an app store rather than on the web can produce a different number, which is why X publishes its own tier prices as "starts at" figures with localized variation.

The third is search and tool usage. Grok's model documentation is explicit that the model has no knowledge of events past its training cut-off unless you enable server-side search tools, and those tools are part of what you are metered on when building. The fourth is currency and tax. xAI and X both publish localized pricing, and the US figures quoted throughout this article exclude VAT and payment-method fees.

None of this changes the headline answer, which remains $30 a month for a serious individual Grok subscription. It changes the answer to "what will this cost my company," which is a different question and needs the API meter, the seat count, and the billing unit nailed down before anyone signs anything.

• Subscription spend and API spend are separate; neither credits the other.
• App-store purchases and non-US currencies can differ from the web prices listed on xAI's page.
• Grok 4.5's knowledge cut-off is 1 February 2026, so anything current requires search tools to be enabled.
• Quoted US figures exclude tax, VAT, and payment-method fees.
• The Grok Business card does not state whether $30 is charged per seat or per workspace.

Is a Grok 4 subscription worth it?

Worth it is a usage question, not a features question. At $30 a month a subscription needs to save you roughly half an hour of work monthly to break even at most professional rates, which daily users clear without noticing and occasional users never reach. Run the sequence below before you subscribe rather than after.

1. 1Spend two weeks on the free tier doing your real work, not test prompts. Same flagship model, so this is a genuine evaluation of answer quality rather than a downgraded demo.
2. 2Count how many times you hit a rate limit or wanted Expert search depth. Fewer than a handful and the free tier is your answer.
3. 3Check whether you already have Grok through X. X Premium at $8 a month includes increased Grok limits, and Premium Business affiliates receive Premium+ with SuperGrok access.
4. 4If you want image and video generation, that alone justifies SuperGrok, since Imagine sits behind the paywall entirely.
5. 5Only consider Heavy if you can name specific questions from the past month where parallel multi-agent reasoning would have changed your answer.
6. 6If your use case is a product rather than personal productivity, skip subscriptions and price the API instead. Software does not need a seat.

What a support agent costs to run instead

Every price in this article is attached to a person. SuperGrok at $30 a month, Grok Business seats, ChatGPT Business at $20 per user: per-seat pricing exists because internal productivity value scales with headcount. Ten employees using an assistant get roughly ten times the value one employee does, so ten times the price is a defensible trade.

Customer support breaks that logic. The value of a support agent scales with the number of customers it handles, not the number of staff who log in. A store with two support people and forty thousand orders a month has a volume problem, not a seat problem, and buying two Grok Business seats does nothing about it because those seats put an assistant in front of your team rather than in front of your customers.

There is also a capability gap that no subscription tier closes. A general-purpose assistant does not know where order 10482 is, cannot check whether a specific item is inside its return window, and has no defined escalation path when it should stop guessing and get a human. Those are integrations and rules, not model quality, which is why every tier listing the same flagship model does not help you here.

Purpose-built support tools price on volume for that reason. Bookbag runs on flat monthly plans with message-credit allowances plus a spend cap you set, rather than per-seat licences. One credit equals one AI reply on any model and a typical conversation runs about four replies, so a plan maps to conversations rather than headcount: Free covers 50 credits, Starter is $30 a month for 600, Growth is $110 for 5,000. Worth avoiding on this side of the market is per-resolution billing, where the invoice grows the better the automation performs, which points the vendor's incentives away from yours.