Grok Pricing Explained (2026): What SuperGrok Really Costs
Grok's pricing page looks simple until you actually try to buy the right thing. There's a free tier, a $10 tier, a $30 tier, a $100 tier and a $300 tier, sold across two different storefronts (grok.com and X), a separate developer API billed per token, per-call tool fees on top of that, and - since June 2026 - a single shared weekly usage pool that spans chat, image generation, voice and coding rather than the old per-product daily caps. Layer on top of all that a corporate rebrand (xAI is now legally SpaceXAI, a wholly-owned AI division of SpaceX) and a $60 billion acquisition of Cursor, with the companies aligning Grok's and Cursor's subscription plans, and "how much does Grok cost" stops being a one-line answer.
This guide prices every current tier from primary sources, explains the shared usage pool mechanic that decides how far your subscription actually goes, and works through what a light, moderate and heavy user should expect to pay once Extra Usage Credits enter the picture.
Quick answer: Grok is free with tight daily limits, $30/month for SuperGrok (the tier most individual users should actually buy), $100/month for the newer SuperGrok Plus, and $300/month for SuperGrok Heavy. The xAI developer API is separate and billed per token — Grok 4.3 remains a lower-priced option at $1.25/$2.50 per million tokens, while Grok 4.7 is the current frontier model at $2/$6 below 200K prompt tokens and $4/$12 at or above that threshold. Whatever plan you're on, once you exhaust your weekly usage allowance you either buy Extra Usage Credits at standard (non-discounted) rates or upgrade — and that gap between "included usage" and "top-up rate" is where a lot of real-world Grok bills quietly grow past the sticker price.
First, the corporate context that changes how you should read every other Grok pricing guide
Two 2026 events matter enough to affect how you plan around Grok, and most pricing roundups don't mention either.
xAI no longer exists as a standalone company. SpaceX completed its all-stock acquisition of xAI in February 2026, and on July 6, 2026 the company formally rebranded from xAI to SpaceXAI. Grok itself did not change names — the chatbot, SuperGrok, and the developer API all keep the Grok brand — but the parent entity, its account handles, and its official documentation now all read "SpaceXAI." If you see docs.x.ai referring to "SpaceXAI Docs," that's the same product, not a different one.
SpaceXAI acquired Cursor (Anysphere) for $60 billion in August 2026, and by mid-September 2026 the company was actively unifying Grok's and Cursor's subscription plans. On September 14, 2026, SpaceXAI's product lead publicly confirmed the two companies are aligning their plans, with completion expected "within several weeks." Right now, Cursor Pro, Pro+ and Ultra all include Grok Bot access, and SuperGrok/X Premium+ can separately enable Grok Bot for Cursor accounts - but the two Grok Bot allowances don't stack, and the subscriptions are still billed separately. This is a genuinely moving target. Treat current Grok-and-Cursor pricing as a snapshot, not a stable structure, and re-check it before standardizing a coding-tool budget around it.
The consumer tiers: what's actually for sale right now
As of September 2026, Grok is sold across two storefronts. Standalone Grok plans are sold directly at grok.com (and the iOS/Android apps); Grok access is also bundled into X's own subscription tiers.
| Plan | Price | Storefront | What it's really for |
|---|---|---|---|
| Free | $0 | grok.com, X | Sampling the product; impractical for sustained use |
| SuperGrok Lite | $10/month | grok.com | Entry-level paid tier, basic Grok Imagine access |
| SuperGrok | $30/month ($300/year) | grok.com | The standard tier for most individual power users |
| SuperGrok Plus | $100/month ($1,000/year) | grok.com | For users who consistently exhaust the $30 tier's usage pool |
| SuperGrok Heavy | $300/month | grok.com | Grok 4 Heavy multi-agent mode, maximum rate limits |
| X Premium | $8/month | X | X platform features first, light Grok access |
| X Premium+ | $40/month ($395/year) | X | Full X perks plus Grok access equivalent to SuperGrok |
The one comparison worth internalizing: SuperGrok at $30/month and X Premium+ at $40/month unlock essentially the same Grok capability — the $10 difference buys you X's ad-free browsing and platform perks, nothing more on the AI side. If you don't personally value the X platform features, SuperGrok is the cheaper door to the exact same Grok access, and choosing X Premium+ instead costs an extra $120/year for no additional AI functionality.
SuperGrok Plus is the newest tier in this lineup, landing in August 2026 specifically to fill the gap between SuperGrok's $30 and Heavy's $300 — it exists because enough users were exhausting the $30 tier's weekly pool without wanting to pay 10x more for Heavy. It adds significantly higher weekly usage, 1080p video generation (up from SuperGrok's standard-definition cap), priority processing, and early feature access.
One documented pricing wrinkle to flag honestly: the public Apple App Store listing shows in-app purchase amounts of $10, $100, $30 and $300 attached to SuperGrok Lite and SuperGrok labels in a way that doesn't map cleanly and unambiguously to a single verified monthly-vs-annual structure for every line item. The clean, confirmed structure above reflects xAI's/SpaceXAI's own FAQ and multiple independently cross-checked pricing trackers as of September 2026 — but as with any subscription, the final checkout screen on your own account and platform (web, iOS, Android) is the actual source of truth, and it can vary by region.
How the shared weekly usage pool actually works
This is the mechanic that decides whether your $30/month actually covers what you need, and it changed significantly in June 2026.
Before June 2026, each Grok product (Chat, Imagine, Voice, Build) had its own separate daily limit. That meant you could run out of image generations while still having plenty of chat messages left, with no way to reallocate between them. Since the rollout, paid Grok users get one shared weekly usage pool they can spend however they like across any combination of Grok products — chat, image/video generation, voice, coding. Different products consume the pool at different rates depending on their compute cost: a text message uses very little of your allowance; generating a high-quality video or running a long coding task uses far more.
According to xAI's/SpaceXAI's own documentation:
- The pool resets weekly, on a schedule shown in your account's Usage settings.
- You can check your current usage percentage and a per-product breakdown (API, Build, Chat, Imagine, Voice) at any time.
- Hitting your weekly limit is not a hard stop. Grok's free-tier chat and voice access remain available on their own separate schedule even after your paid weekly pool is exhausted. Beyond that, you have three options: buy Extra Usage Credits, upgrade to a higher plan, or turn on Auto Top Up.
Extra Usage Credits: the part of the bill most guides skip
Once your included weekly usage runs out, you can buy Extra Usage Credits to keep going immediately — as little as $5 at a time, purchasable only on the web (in-app mobile purchases aren't yet supported). Credits land in your account instantly, don't expire for a year after purchase, and are consumed after your plan's included weekly usage runs dry.
The detail that matters for a real budget: Extra Usage Credits are priced at standard rates, meaning the effective cost per action is meaningfully higher than what you're getting from your plan's included weekly allowance. xAI's own FAQ states this plainly — "upgrading to a higher plan provides significantly more weekly usage at a lower cost per action compared to repeatedly buying Extra Usage Credits." In practice, that means:
- Occasional overage (you blow past your limit once or twice a month): buying $5–$10 of top-up credits is convenient and cheap in absolute terms.
- Regular overage (you're buying credits most weeks): you're paying a premium for usage that a higher subscription tier would have covered more cheaply per action. This is the single most common way a Grok bill quietly drifts from "$30/month" to something closer to $50–$80/month without the user ever consciously deciding to upgrade.
Auto Top Up lets you set an automatic credit purchase (with your own chosen amount and monthly cap) so usage never interrupts — convenient, but worth setting a conservative monthly cap given the standard-rate pricing on those credits.
Grok Imagine: image and video generation costs
Grok's image/video generation feature has had one of the more volatile access histories of any major AI product in 2026, and the free tier has effectively been closed off entirely.
Free-tier image generation was removed on March 19, 2026. As of that date, free accounts cannot generate images or video through Grok Imagine at all — access requires a paid SuperGrok tier (or the API). This is a meaningful downgrade from Grok Imagine's original 2026 launch, when it briefly offered generous free access and generated over a billion videos in its first month.
Generation limits on paid tiers have also tightened materially over the course of the year. Independent testing and xAI support communications reported standard SuperGrok capped around 20+ videos per 24 hours in May 2026 (down sharply from earlier, more generous per-12-hour allowances), with SuperGrok Heavy users getting a higher but still finite allowance. Since the June 2026 shared-pool rollout, these separate daily video/image caps have been folded into the same weekly usage pool as everything else — so the practical limit today is "how much of your weekly pool you're willing to spend on video versus chat versus coding," rather than a fixed per-day video count.
Resolution is also tier-gated. In the consumer app, you choose clip length (6, 10 or 15 seconds) and resolution (480p or 720p) in the composer, but your account tier decides which options actually run — a 720p request will silently fall back to 480p once you hit your tier's 720p cap. SuperGrok Plus specifically adds 1080p generation, which lower tiers don't offer at all.
Watermarking is not removable on any tier. Every generated image and video carries a visible "grok" watermark, both to disclose AI generation and, in some jurisdictions, to comply with legal labeling requirements — there's no setting on any paid plan to remove it.
Through the API, grok-imagine-video-1.5 is priced and controlled differently: it accepts any whole-second length from 1–15 seconds, reaches 1080p on image-to-video generation specifically, and reference-to-video tops out at 720p — giving developers more granular control than the consumer app's fixed length/resolution buttons.
The xAI/SpaceXAI developer API: full pricing table
The developer API is a completely separate product from any consumer subscription — subscribing to SuperGrok does not include API credits, and API usage is billed independently per token. All figures below are per 1 million tokens, sourced from SpaceXAI's own developer documentation and current as of September 25, 2026.
| Model | Input (per 1M) | Cached input (per 1M) | Output (per 1M) | Context |
|---|---|---|---|---|
| Grok 4.7 (current frontier model) | $2.00 | $0.50 | $6.00 | 500K |
| Grok 4.6 (previous generation) | $2.00 | $0.50 | $6.00 | 500K |
| Grok 4.5 (previous generation) | $2.00 | $0.30 | $6.00 | 500K |
| Grok 4.3 (lower-priced 1M-context option) | $1.25 | — | $2.50 | 1M |
| Grok 4.20 variants (reasoning/non-reasoning/multi-agent) | $1.25 | — | $2.50 | 1M |
| Grok Build 0.1 (coding, cheapest tier) | $1.00 | — | $2.00 | 256K |
The 200K-token cliff. Grok 4.7 uses a threshold, not a slope: once a single request's prompt reaches 200,000 tokens, the entire request is billed at double the standard rate — $4.00 input / $1.00 cached / $12.00 output per million tokens — not just the tokens above the line. This is a meaningful trap for long-document or large-codebase workflows that occasionally cross the threshold without careful token budgeting; a request at 199K tokens and one at 201K tokens can cost roughly twice as much per token for the larger one, even though the difference in actual content is trivial. The same threshold pricing applies to Grok 4.6 and 4.5 where those models remain available.
Tool calls are billed separately from tokens. Server-side tools carry their own per-call fees on top of whatever tokens the model consumes: web search, X search and code execution are billed at roughly $5 per 1,000 calls each; file attachment processing at roughly $10 per 1,000 calls; collections search at roughly $2.50 per 1,000 calls. If your application makes heavy use of tool calls (an agent that searches the web multiple times per user query, for instance), these per-call fees can add up faster than the token cost itself on short, tool-heavy interactions.
Legacy models silently redirect, and still bill at the new rate. On May 15, 2026, SpaceXAI retired the original Grok 4, Grok 4 Fast, Grok 4.1, Grok 4.1 Fast, Grok Code Fast 1 and Grok 3 model IDs. Requests to these old model names don't fail — they're silently routed to Grok 4.3 (or Grok Build 0.1 for the coding-specific IDs) and billed at the current rate for the replacement model, even in cases where the original model was actually cheaper. If your codebase has a hardcoded model string from earlier in 2026, it's worth an explicit audit: your requests are still succeeding, but your bill may no longer match the price you originally budgeted for.
Real cost scenarios
These are SCENARIO / PLANNING ASSUMPTIONS, not figures published by SpaceXAI as typical or guaranteed spend.
Light user: daily chat questions, no image generation
Sticking to the free tier is workable if your needs are genuinely occasional — SpaceXAI doesn't publish an exact daily prompt cap, and independent testers have reported figures in the range of roughly 10 prompts per rolling window, with a smaller context window than paid tiers. For anything beyond casual daily use, the free tier's limits bite quickly. Cost: $0/month, with the caveat that "free" here means genuinely light, occasional use.
Moderate user: daily driver replacing another $20/month subscription
A user who wants Grok as their primary assistant — daily chat, occasional image generation, DeepSearch for research — fits comfortably on SuperGrok at $30/month ($300/year). That's $10 more than ChatGPT Plus or Claude Pro at $20/month, a gap worth being clear-eyed about: you're paying a 50% premium over the Western consumer-subscription default price point for Grok's specific differentiators (real-time X data access, DeepSearch, the model's particular reasoning/personality profile). Whether that premium is worth it depends entirely on whether those specific features matter to your workflow — for a user who never touches X data or DeepSearch, the $10/month gap buys little that a $20/month competitor doesn't already offer.
Heavy user: hits weekly limits regularly, buys top-ups most months
A user on the $30 SuperGrok tier who regularly exhausts their weekly pool and buys, say, $15/month in Extra Usage Credits at standard (non-discounted) rates to keep going is effectively paying $45/month for SuperGrok-level access — 50% more than the sticker price, and (per SpaceXAI's own guidance) less efficient per action than simply upgrading. The math favors moving to SuperGrok Plus at $100/month once top-up spending becomes a monthly habit rather than an occasional overflow, since Plus's larger included pool covers materially more usage per dollar than repeated top-ups on the $30 tier.
Power user / developer: API-based automated workflow
A small internal tool processing roughly 20 million input and 5 million output tokens a month on Grok 4.3 (1M context, no long-context penalty triggered): 20M × $1.25 + 5M × $2.50 = $25.00 + $12.50 = $37.50/month in raw token cost, before any tool-call fees. Switching the same volume to the coding-optimized Grok Build 0.1 (assuming the task fits its 256K context): 20M × $1.00 + 5M × $2.00 = $20.00 + $10.00 = $30.00/month — a roughly 20% saving for workloads that fit comfortably under Build's smaller context ceiling.
Grok vs. ChatGPT Plus and Claude Pro: does the $10 premium replace anything?
SuperGrok's $30/month sits $10 above ChatGPT Plus, Claude Pro and Google's AI Pro tier, all clustered at $20/month. The bundling logic that makes NotCheapAI's methodology matter here: does SuperGrok replace another subscription you'd otherwise be paying for, or does it sit alongside them as an additional cost? For most users evaluated honestly, Grok doesn't yet fully substitute for a general-purpose assistant subscription the way, say, a coding-specific tool might replace a separate general-purpose one — it's genuinely additive for most buyers, which means the real annual comparison isn't $360 (SuperGrok) vs. $240 (ChatGPT Plus), it's whether you're willing to run both at a combined $600+/year, or pick one based on which specific features (X data access and DeepSearch vs. a broader third-party plugin/connector ecosystem) matter more to your actual use case.
The API-level comparison is more clearly favorable to Grok: Grok 4.3 at $1.25/$2.50 per million tokens with a 1M context window remains the lower-priced option, while current Grok 4.7 at $2/$6 below 200K prompt tokens roughly matches Claude Sonnet 5's input price and comes in meaningfully cheaper on output. At or above 200K prompt tokens, Grok 4.7's entire request is charged at $4/$12 per million input/output tokens, so the comparison changes for long-context workloads.
Hidden costs and things that make the sticker price misleading
- The $10/$30/$100/$300 ladder isn't the whole story once Extra Usage Credits enter the picture. A "cheap" $30/month plan that requires regular top-ups at standard, non-discounted rates can quietly cost as much as the next tier up, while being less efficient per action.
- The 200K-token cliff on Grok 4.7 doubles your entire bill for a request, not just the excess. Workloads that occasionally send long documents need explicit token budgeting to avoid crossing this line unintentionally; the same threshold applies to 4.6 and 4.5.
- Tool calls bill separately from tokens. An agent that leans heavily on web search or file attachments can rack up per-call fees that dwarf the underlying token cost on short interactions.
- Legacy model IDs silently redirect and rebill at current rates. If you built against Grok 4 or Grok 4.1 Fast earlier in 2026, your integration still works — but it's no longer charging the price you originally tested against.
- Grok-and-Cursor pricing is actively changing. SpaceXAI confirmed on September 14, 2026 that it's unifying Grok's and Cursor's subscription structures within weeks. Anyone budgeting a coding-tool stack around today's Cursor-plus-Grok-Bot arrangement should expect that structure to change imminently.
- X-bundled plans show up on statements as X subscriptions, not AI tool spend. If your expense reporting or budget categorization matters, the standalone SuperGrok tiers are cleaner to track than X Premium/Premium+.
Is SuperGrok worth paying for in 2026?
For someone who wants Grok specifically — its particular model behavior, its real-time X data access, DeepSearch — and will use it daily, SuperGrok at $30/month is the right entry point, not X Premium+ at $40, and not SuperGrok Lite at $10 (which caps out on image generation quickly). For someone who regularly exhausts that tier's weekly pool, SuperGrok Plus at $100/month is a more honest price than repeatedly buying standard-rate top-up credits. SuperGrok Heavy's $300/month is a genuinely narrow tier: it earns its keep only for users doing sustained, compute-heavy multi-agent reasoning work where Grok 4 Heavy's capability difference matters, not as a default upgrade path from Plus. For developers, Grok 4.3 remains the lower-priced API option; Grok 4.7 is the current frontier model, with the 200K threshold materially increasing costs for long-context requests.
FAQ
How much does SuperGrok cost? $30/month or $300/year for the standard SuperGrok tier. A newer SuperGrok Plus tier sits above it at $100/month, and SuperGrok Heavy sits at the top at $300/month. A lower SuperGrok Lite entry tier is available at $10/month.
What's the difference between SuperGrok and X Premium+? They unlock essentially the same Grok access. X Premium+ at $40/month additionally bundles X platform perks (ad-free browsing, checkmark, publishing features); SuperGrok at $30/month is the cheaper standalone path if you don't specifically want those X features.
What happens when I run out of my weekly Grok usage? Your paid features pause until the weekly pool resets, but you're not locked out entirely — Grok's free-tier chat and voice limits remain available on their own schedule. You can also buy Extra Usage Credits (from $5), upgrade your plan, or enable Auto Top Up to continue immediately.
Are Extra Usage Credits a good deal? For occasional overage, yes — buying a small top-up once or twice a month is cheap and convenient. For regular overage, no — SpaceXAI's own documentation states that credits are priced at standard rates, making them less cost-efficient per action than simply upgrading to a plan with a larger included weekly pool.
How much does the Grok API cost? Grok 4.7 is the current frontier API model (released September 21, 2026): $2/$0.50/$6 per million input/cached-input/output tokens below 200K prompt tokens, and $4/$1/$12 at or above 200K. Grok 4.3 remains a lower-priced 1M-context option at $1.25/$2.50 per million input/output tokens. Grok Build 0.1, the coding-specific model, is listed at $1.00/$2.00 per million tokens in the pricing comparison used here.
Can I still generate images for free on Grok? No. Free-tier image and video generation through Grok Imagine was removed on March 19, 2026. Generation now requires a paid SuperGrok tier or API access.
Is xAI the same company as SpaceXAI? Yes. SpaceX acquired xAI in an all-stock deal that closed in February 2026, and xAI formally rebranded to SpaceXAI on July 6, 2026. Grok, SuperGrok and the developer API kept their existing names; only the parent corporate brand changed.
Does buying SuperGrok include access to Cursor or Grok Bot? Partially, and this is actively changing. As of September 2026, SuperGrok and X Premium+ can enable Grok Bot access for Cursor accounts, and Cursor's own Pro/Pro+/Ultra plans separately include Grok Bot — but the two allowances don't stack, and the subscriptions remain billed separately. SpaceXAI has publicly confirmed it is unifying these plans, with completion expected within several weeks.
What is the 200K-token pricing cliff? On Grok 4.7, once a single request's prompt reaches 200,000 tokens, the entire request — not just the tokens beyond the threshold — is billed at double the standard rate. This makes occasional long-document requests significantly more expensive than their token count alone would suggest. Grok 4.6 and 4.5 also use this threshold pricing.
Do old Grok model names still work? Yes, but they bill differently than you might expect. Legacy model IDs like grok-4, grok-4-fast and grok-code-fast-1 (retired May 15, 2026) silently redirect to grok-4.3 or grok-build-0.1 and are billed at that replacement model's current rate, which in some cases is different from the original model's price.