SuperGrok Heavy: what $300 a month actually buys

SuperGrok Heavy is the most expensive consumer AI subscription on the market at $300 a month, and the pages describing it mostly repeat two lines from a launch post. Two facts change the arithmetic. Heavy is only $100 above Cursor Ultra, which grants the same agent entitlement, so if you were buying Ultra anyway the real increment is small. And Heavy includes no API credit at all, so the workload most people imagine justifying it is billed separately.

By the Continuum team. We build a workbench that runs Claude Code, Codex, and their peers, so the model rates quoted here are the ones our own cost analytics ship with.

The short version

SuperGrok Heavy is the top consumer tier of xAI’s subscription ladder, reported at $300 a month, which is $3,600 a year with no annual discount published. It buys three things nothing cheaper does: Grok 4 Heavy, a multi-agent mode where a leader agent coordinates up to 16 sub-agents on one problem, the highest Grok Bot allowance of any plan, and the highest usage ceilings across chat, Imagine, voice, and Grok Build. It does not buy API credits, an enterprise agreement, a spend cap on Grok Bot, or a model picker. The comparison that matters is Cursor Ultra at $200, which also grants Grok Bot: measured against a plan you may already want, Heavy’s marginal cost is about $100 a month rather than $300. A promotional price of $99 a month is circulating, with sources disagreeing on whether it runs three months or six.

What you need to know
  • $300 a month, $3,600 a year. No annual discount for Heavy is reported anywhere, unlike SuperGrok and SuperGrok Plus which both offer roughly two months free.
  • The headline feature is Grok 4 Heavy: a leader agent coordinating up to 16 sub-agents on one problem, then synthesising their work.
  • Heavy includes no API credits. The xAI API is a separate self-serve product billed per token, and no subscription tier feeds it.
  • The real increment is about $100. Cursor Ultra at $200 also grants Grok Bot, so Heavy is $100 above a plan many buyers already want.
  • Heavy’s Grok Bot allowance is the highest available, which matters because that allowance is unpublished and its overage is uncapped.
  • A $99 promotion is reported but unverified, at three months by one source and six by another. A rumour to check at checkout, not a price to plan on.

What SuperGrok Heavy is

SuperGrok Heavy is the top rung of xAI’s consumer subscription ladder, sitting above SuperGrok Plus at $100 and SuperGrok at $30. It is reported at $300 a month everywhere that names a figure, and published by xAI on no page we could load: both x.ai/pricing and grok.com/pricing returned HTTP 403 to automated fetching on 22 August 2026. Treat the price as reported rather than published and confirm it at checkout.

It buys four things, and only one of them is exclusive to the tier.

What Heavy adds over the tier below it.checked 22 aug 2026
What Heavy addsExclusive to Heavy?What it means in practice
Grok 4 Heavy, the multi-agent modeYesA leader agent runs up to 16 sub-agents on one problem, then synthesises. Nothing cheaper carries it
The highest Grok Bot allowanceYes, in degreePlus and the Cursor tiers also grant Grok Bot; Heavy grants the most of it. Cursor’s docs put Plus "below the Heavy tier"
Highest usage ceilings elsewhereYes, in degreeChat, Imagine, voice, and Grok Build all run at the top limits
Priority access to new modelsNoPlus carries priority too. This is a difference of position in a queue, not of entitlement

The 16-agent multi-agent mode

Grok 4 Heavy is a multi-agent architecture rather than a bigger model. The shape, as xAI documents it on the API side of the same capability, is a leader agent that dispatches work to sub-agents, each with its own focused context, then synthesises their outputs into one answer. The API exposes the same idea as a model with a configurable agent count: 4 agents for a quick pass, 16 for a deep one.

Three consequences follow, and they pull in different directions.

01

It genuinely helps on wide problems

Research sweeps, "check this claim against every source you can find", exhaustive option comparisons, anything where the bottleneck is breadth of exploration rather than depth of a single chain of thought. Sixteen contexts covering sixteen angles beats one context trying to hold all sixteen.

02

It costs sixteen times the tokens

On the API, xAI is explicit that every leader and sub-agent token is billed, including reasoning tokens and each agent’s tool calls. Nothing about the subscription version changes the underlying economics; it changes who absorbs them. On a $300 plan you absorb them as allowance rather than as invoice, which is exactly why the tier costs what it does.

03

It does not help on narrow problems

A single well-scoped code change, a focused debugging session, a specific question with a specific answer. Sixteen agents exploring one narrow problem produce sixteen versions of the same finding and then a synthesis step that adds latency. The mode is a tool with a shape, not a general upgrade.

The honest framing is that Grok 4 Heavy is a breadth purchase. If your hardest weekly task is "find everything relevant about X and reconcile it", the mode is a real capability and the tier is the only way to get it on a subscription. If your hardest weekly task is a difficult code change, you are paying $270 above the $30 tier for something that will not help with it, and Grok Build at $30 will.

What Heavy does not buy

This is the section most Heavy write-ups skip, and it contains the assumption that costs people the most.

  • No API credits, and no API relationship at all. The xAI API is a separate self-serve product with its own billing at console.x.ai. A $300 subscription contributes nothing to it, and API spend does not count against the subscription. If your plan for Heavy involves scripting anything against xAI models, that is a second bill.
  • No spend cap on Grok Bot. Heavy grants the highest agent allowance, whose size xAI does not publish, and past it overage bills at raw model rates with no cap available. The most expensive plan still has an uncapped meter underneath it.
  • No model picker in Grok Bot. Routing is automatic and undisclosed, so you cannot slow the meter by sending cheap work to a cheaper model.
  • No enterprise agreement. Heavy is a consumer plan. There is no DPA, no support obligation, and no audit surface that comes with it. The business plans, SuperGrok Team and SuperGrok Enterprise, are a different product, and per Cursor’s documentation they do not grant Grok Bot at all.
  • No escape from the Cursor link. Grok Bot access requires a Cursor account, and Cursor’s documentation states the link is permanent: "You can’t unlink a Grok account from a Cursor account, and you can’t move a link to a different Cursor account."
  • No annual discount, as far as anyone reports. SuperGrok and SuperGrok Plus both offer roughly two months free on annual. Heavy is quoted monthly everywhere.

The Cursor Ultra arbitrage

This is the calculation that most changes whether Heavy is a sensible purchase, and it appears in roughly one place on the internet.

Grok Bot is granted by six plans, and two of them sit at the top of their respective ladders: SuperGrok Heavy at a reported $300, and Cursor Ultra at $200. If you are a developer who was already considering Cursor Ultra, then the agent entitlement is not something Heavy uniquely provides. Measured that way, Heavy costs $100 more than a plan you already wanted, and that $100 buys Grok 4 Heavy plus the higher allowances.

Heavy priced against the alternatives that carry the same agent entitlement. Prices reported, not read from a vendor page.
PathMonthlyAnnualGrok BotGrok 4 HeavyCursor editor
SuperGrok ($30) only$30$300NoNoNo
Cursor Pro+$60Published 20% off annualYesNoYes
SuperGrok Plus$100$1,000YesNoNo
Cursor Ultra$200Published 20% off annualYesNoYes
SuperGrok Heavy$300Not publishedYes, highestYesNo
SuperGrok ($30) + Cursor Ultra$230MixedYesNoYes
Heavy + Cursor Pro ($20)$320MixedYes, highestYesYes

Three readings of that table, in order of how often they apply.

  • If you would buy Cursor Ultra anyway, Heavy’s effective increment is about $100 a month. That is a very different decision from a $300 one, and it is the only framing under which Heavy is an easy yes for an individual developer.
  • If you would not buy Cursor at all, the honest comparison is $300 against $100. SuperGrok Plus grants Grok Bot too. The $200 difference buys the multi-agent mode and more allowance, and nothing else.
  • If you want the Cursor editor as well, note that Heavy does not include it. Heavy grants Grok Bot through a required Cursor account, but the account it grants you is not a paid Cursor editor plan. The combined row is $320, not $300.

The reported $99 promotion

A promotional price cutting Heavy from $300 to $99, a 67 per cent discount, is circulating in the trade press. We could not verify it against xAI, and the reporting on its duration conflicts, so treat it as a rumour to check at checkout rather than a price to plan on. The divergence is worth $600 over a year.

What different sources say about the same promotion. Low confidence throughout.
Reported termsCost over the windowThen
$99 a month for six months$594 against $1,800 at listReverts to $300
$99 a month for three months, targeted offer flow$297 against $900 at listReverts to $300
A stacked full-year figure claimed by promotion-tracking accountsUnverifiedUnknown

Two things to do with that. Budget for three months and treat six as upside, because the failure mode of the optimistic assumption costs you $600 and the failure mode of the pessimistic one costs you nothing. And decide whether Heavy is worth $300, not whether it is worth $99, because $300 is the price in month four. A tier that is only worth buying on promotion is a tier you should leave when the promotion ends, which means the real question is whether you will actually cancel.

Is Heavy worth it, by user type

The tier is genuinely right for a narrow set of people and genuinely wrong for a much larger one. Rather than a verdict, here is the sorting.

Heavy at $300 a month, assessed against what each kind of buyer is actually trying to do.
You areHeavy?Why, and what to buy instead
A researcher or analyst whose hard tasks are breadth-shapedYesThe 16-agent mode is a real capability for exhaustive sweeps and nothing cheaper carries it
Already buying Cursor Ultra at $200ProbablyThe increment is about $100 for multi-agent plus the top allowances
Running Grok Bot on real recurring work, dailyProbablyHeavy has the highest allowance, and the allowance is the uncapped part. Higher headroom is worth real money here
A developer who wants an xAI coding agentNoSuperGrok at $30 carries Grok Build. That is a tenth of the price for the thing you want
Someone who wants Grok Bot and nothing elseNoCursor Pro+ at $60, or a Cursor Teams Standard seat at $40
Scripting against Grok modelsNoThe API is self-serve, per token, and no subscription feeds it. Heavy adds a bill without reducing one
A team needing SSO, audit, or a DPANoHeavy is a consumer plan. Business plans exist and, per Cursor’s docs, do not grant Grok Bot
Evaluating always-on agents in generalNoEvaluate on the cheapest qualifying plan with an account you would not mind stranding, since the Cursor link is permanent

The pattern across those rows: Heavy is a headroom purchase, and headroom is only worth $3,600 a year if you have already hit the ceiling below it. Almost nobody buying Heavy as a first Grok subscription has. The defensible sequence is $30, then $100 if agents start to matter, then $300 only when you can point at the specific week where Plus ran out.

The alternative nobody prices against it

$3,600 a year is a meaningful budget, and it is worth seeing what else it reaches. The comparison is not to a cheaper subscription; it is to a different shape of spend.

The xAI API prices grok-4.6 at $2 per million input tokens, $0.50 cached, and $6 per million output for 200,000 prompt tokens or fewer, doubling to $4 and $12 above 200,000, with the higher rate applying to the whole request rather than the excess. At those rates, Heavy’s annual cost buys roughly 600 million output tokens, or considerably more if your work caches well. Whether that is more or less than Heavy grants you is unanswerable, because xAI does not publish the allowance, and that asymmetry is itself the argument: one side of this comparison has a published unit price and the other does not.

That is not an argument that nobody should buy Heavy. It is an argument that the thing Heavy is really selling, at $300 a month, is the removal of a decision: no keys, no metering, no model choice, no machine to keep awake. Some people should buy that and should not feel bad about it. The people who should not are the ones who bought it to reduce a bill it does not touch, or to get a coding agent that costs $30.

Questions people ask

What is SuperGrok Heavy?

SuperGrok Heavy is the top consumer tier of xAI’s subscription ladder, reported at $300 a month. It grants Grok 4 Heavy, a multi-agent mode where a leader agent coordinates up to 16 sub-agents on one problem and synthesises their work, the highest Grok Bot allowance of any plan, and the highest usage ceilings across chat, Imagine, voice, and Grok Build. It is the only tier that carries the multi-agent mode.

How much is SuperGrok Heavy?

A reported $300 a month (xAI publishes no Heavy price on any page we could load, with both x.ai/pricing and grok.com/pricing returning HTTP 403 on 22 August 2026), which is $3,600 a year. No annual discount is reported for Heavy, unlike SuperGrok at $300 a year and SuperGrok Plus at $1,000. A promotional price of $99 a month is circulating, with sources disagreeing on whether it lasts three months or six, and we could not verify that either.

Is SuperGrok Heavy worth it?

At a reported $300 a month, it is worth it if the multi-agent mode solves a problem you actually have, if you were buying Cursor Ultra at $200 anyway so the real increment is about $100, or if you are running Grok Bot on daily recurring work and need the highest allowance. It is not worth it if you want a coding agent, since SuperGrok at $30 carries Grok Build, or if you want Grok Bot alone, since Cursor Pro+ at $60 grants it.

What is the difference between SuperGrok Plus and SuperGrok Heavy?

Both grant Grok Bot. Plus, at a reported $100, grants what Cursor’s documentation calls generous usage "below the Heavy tier"; Heavy, at a reported $300, grants the highest allowance available. Neither price could be read from an xAI page, both of which return HTTP 403 to automated fetching. The exclusive difference is Grok 4 Heavy, the 16-agent multi-agent mode, which Plus does not include. Everything else between the two tiers is a difference of quantity rather than capability.

Does SuperGrok Heavy include API credits?

No. The xAI API is a separate, self-serve product billed per token at console.x.ai, and no subscription tier contributes credit to it. API spend also does not count against a subscription allowance. If your workload runs through api.x.ai, Heavy adds a $300 monthly bill without reducing the one you already have.

Does SuperGrok Heavy include Cursor?

It grants Grok Bot, which requires a Cursor account to use, but it does not grant a paid Cursor editor plan. If you want the Cursor editor as well, price it as a separate line: Heavy plus Cursor Pro is $320 a month. Note also that Cursor Ultra at $200 grants Grok Bot on its own, which is the comparison that makes Heavy look expensive or cheap depending on what you were already buying.

How many agents does Grok 4 Heavy run?

Up to 16. xAI’s API documentation for the same capability describes a leader agent dispatching work to sub-agents, each with its own focused context, configurable to 4 agents for a quick pass or 16 for a deep one, with the leader synthesising the results. On the API every leader and sub-agent token is billed, including reasoning tokens and each agent’s tool calls, which is a useful indication of what the subscription version is absorbing on your behalf.

Can I get SuperGrok Heavy for $99?

A $99 promotion is reported, but the reporting conflicts and we could not verify it against xAI, so treat it as a rumour to check at checkout rather than a price to plan on. One outlet reports six months, a first-hand account describes a targeted three-month flow. If it is live on your account, budget for three months and treat six as upside, because the optimistic assumption costs $600 if it is wrong. Whatever the checkout screen states at the moment you subscribe is the only figure that binds.

Is there a SuperGrok Heavy annual plan?

None is reported. SuperGrok and SuperGrok Plus both have annual pricing at roughly two months free, but every source that names a Heavy price names a monthly one. If an annual Heavy option exists at checkout it is undocumented, and the current promotional activity suggests xAI is discounting Heavy with time-boxed offers rather than an annual commitment.

Does SuperGrok Heavy give me an enterprise agreement?

No. Heavy is a consumer plan with no DPA, no support obligation, and no audit surface. The business tiers are sold separately as SuperGrok Team and SuperGrok Enterprise, and Cursor’s documentation states that those plans do not grant Grok Bot usage at all, which is the opposite of what most buyers expect from an upgrade.

Sources

Every figure above was read from these pages on August 2026. Vendors reprice without notice; if you find a stale number, tell us.

  1. Cursor: SuperGrok Heavy and Grok Bot which SuperGrok plans grant Grok Bot, the Team and Enterprise exclusion, and the permanent account link
  2. Cursor pricing Hobby, Pro, Pro+, Ultra, Teams Standard and Premium, and which of them carry Grok Bot
  3. Grok Bot get started, docs.x.ai the eligible-plan list for Grok Bot and the Cursor sign-in requirement
  4. xAI multi-agent model capabilities the leader plus sub-agent design, the 4 and 16 agent settings, and the billing rule
  5. xAI model list and rates, docs.x.ai published per-million token rates and the 200K context pricing tier
  6. xAI pricing (403 to automated fetch) the canonical tier list; returned HTTP 403 to automated fetching on 22 August 2026
  7. TestingCatalog on the Heavy price promotion the reported 67 per cent Heavy discount; duration conflicts with other reports
  8. Continuum pricing the free desktop app, and Plus, Max 100, Max 200, and Ultra hosted inference
Try it

Sixteen agents.
Your own machine.

Continuum runs Claude Code, Codex, Cursor, Grok, Gemini, and OpenCode in parallel worktrees on hardware you already own, with live quota gauges and spend by repository. Free, with hosted inference if you want it.

free app · your subscriptions · local-first