NotCheapMAKE EVERY CREDIT COUNT

Grok / SuperGrok vs ChatGPT 2026: Real Cost

DIRECT ANSWER

Short Answer

xAI now publicly brands as SpaceXAI — confirmed directly on its own pricing and developer-docs pages ("© 2026 SpaceXAI LLC," "SpaceXAI Docs"). Most existing coverage still calls the company "xAI," which remains the common name for the product and API.

Consumer pricing: xAI's pricing page lists Free ($0/month) with real-time web/X search, voice mode, connectors and SOC 2 compliance; SuperGrok ($30/month) with Grok 4.6, Grok Bot access, higher rate limits, "Expert" search, image and video generation; and SuperGrok Plus ($100/month) with everything in SuperGrok plus 1080p video, "significantly higher usage" across Chat/Imagine/Voice/Build, priority access and early feature access. The page lists SuperGrok Lite and Heavy but does not show their prices in the plan details; independent trackers reported $10/month (Lite) and $300/month (Heavy) through mid-September 2026. Business is reported at $30/seat/month.

A version difference appears between xAI's two pricing surfaces: the consumer plan page describes SuperGrok's model as Grok 4.6, while the developer docs page (last updated September 21, 2026) lists Grok 4.7 as the flagship API model. Public materials do not establish whether SuperGrok consumer users already have 4.7 access or remain on 4.6.

API pricing, confirmed directly on docs.x.ai: Grok 4.7 — $2.00 per million input tokens, $6.00 per million output tokens, 500K context window. Voice API: $0.08/minute (agent), $15.00 per million characters (TTS), $0.10/hour (batch STT), $0.20/hour (streaming STT). Imagine API: images starting at $0.02 each (1K/2K), video starting at $0.05/second (480p/720p/1080p). SuperGrok subscriptions do not include API credits — the consumer app and the developer API are billed entirely separately, the same structural pattern seen across many AI products that separate consumer subscriptions from developer API billing.

Against ChatGPT, at three normalized workloads (10M/2M, 100M/20M, 1B/200M input/output tokens): Grok 4.7 costs $32 / $320 / $3,200, versus GPT-5.6 Sol's $80 / $800 / $8,000 (CALCULATION, using Sol's confirmed $4/$20 API rate) — Grok's flagship API tier is 2.5× cheaper per token than OpenAI's comparable flagship. On the consumer side, SuperGrok ($30/month) costs 50% more than ChatGPT Plus ($20/month) for a broadly comparable "full model access, higher limits" tier.

Current Pricing / Product Map

SurfacePlanPrice
grok.comFree$0/mo
grok.comSuperGrok Lite$10/mo (reported)
grok.comSuperGrok$30/mo
grok.comSuperGrok Plus$100/mo
grok.comSuperGrok Heavy$300/mo (reported; one source listed a $99 new-subscriber offer)
grok.comBusiness$30/seat/mo (reported)
grok.comEnterpriseCustom
X platformX Premium~$8/mo (reported)
X platformX Premium+~$40/mo (reported; feature equivalence with standalone SuperGrok is not established)
docs.x.aiGrok 4.7 API$2.00/$6.00 per 1M tokens
docs.x.aiVoice, Imagine APIssee rates above

Annual billing: reports conflict. One lists annual prices of $100/$300/$1,000/$3,000 for Lite/SuperGrok/Plus/Heavy; another says the plans are month-to-month with no annual discount. Check the checkout page for the current billing option.

Model retirement: xAI's migration guide, dated May 15, 2026, says the original Grok 4, Grok 4 Fast, Grok 4.1, Grok 4.1 Fast, Grok Code Fast 1 and Grok 3 were retired; requests to those model IDs now silently redirect to Grok 4.3 or newer and bill at the new model's rate. This can change costs for code that still pins a retired model ID.

Normalization Method

MODELING ASSUMPTION: normalized at three volumes chosen for cross-provider comparability — 10M input + 2M output tokens/month, 100M input + 20M output, 1B input + 200M output. This is a token-cost comparison for the API only; it does not attempt to normalize consumer-subscription usage, since xAI, like most consumer AI chat products, does not publish a per-message or per-task rate for SuperGrok.

Workload Tables

API cost at normalized volumes (CALCULATION):

WorkloadGrok 4.7GPT-5.6 Sol (reference)
10M in + 2M out$32.00$80.00
100M in + 20M out$320.00$800.00
1B in + 200M out$3,200.00$8,000.00

Consumer subscription comparison:

PlanPriceRough ChatGPT equivalent
Grok Free$0ChatGPT Free
SuperGrok$30/moChatGPT Plus ($20/mo) — SuperGrok costs 50% more
SuperGrok Plus$100/moChatGPT Pro's $100 tier — same headline price
SuperGrok Heavy$300/mo (reported)ChatGPT Pro's $200 tier - Heavy costs 50% more at list, though a reported $99 promotional rate would make it cheaper

Hidden Costs / Limits

  • xAI does not publish how large each plan's weekly usage pool actually is (one source notes this explicitly) — SuperGrok's "higher rate limits" and SuperGrok Plus's "significantly higher usage" are relative descriptions, not numbers, so neither can be converted into a message count the way this article could for some products with published usage tiers.
  • Free tier lost image/video generation in 2026 — multiple convergent sources report xAI removed free-tier Grok Imagine access (one source dates this to March 19, 2026) to control compute costs; free users get text/voice/search only now.
  • Model-ID redirects bill silently at the new rate — the May 15, 2026 retirement of five model IDs means any code still targeting the old names does not fail, it just costs a different amount than budgeted.
  • Tool calls bill separately from tokens. One detailed report lists web search, X search and code execution at $5 per 1,000 calls each; xAI's public API pricing information cited here does not itemize those charges.
  • Grok Bot (an agentic feature) is reported to have its own usage pool, separate from the standard chat allowance, following an August 26, 2026 change.
  • X Premium+ pricing is reported inconsistently across sources ($350 vs. $395/year for one variant) — flagged as unresolved rather than picked.
  • Unspecified geographic restrictions. xAI says model access "might vary depending on various factors such as geographical location, account limitations, etc." but does not enumerate specific blocked countries in the cited information.

Commercial / International Caveats

Training-data opt-out is plan-gated: independent reports say Free and standard consumer conversations may be used to improve xAI's models, while Business and Enterprise plans guarantee exclusion from training. The cited pricing grid lists "No training," "Custom data retention," SSO, SCIM and customer-managed encryption keys as Business/Enterprise features; it does not clearly state the consumer-plan data-use policy. The grid also lists SOC 2 compliance (Type I & II) across paid tiers.

BUILD & EARN Scenario: A Research-Summary API Wrapper

MODELING ASSUMPTION: a service charges $0.02 per summarized document (3,000 input + 300 output tokens per doc), processing 500,000 documents/month = $10,000 revenue, the same scenario shape used across comparable workload models. That workload is 1.5B input + 150M output tokens/month.

CALCULATION: 1,500 × $2.00 + 150 × $6.00 = $3,900.00 in Grok 4.7 token cost. Platform-cost share of $10,000: 39.0%; model-cost gross margin: 61.0%. For comparison, the same workload on GPT-5.6 Sol ($4/$20) costs 1,500×$4 + 150×$20 = $9,000.00 (90.0% of revenue, 10.0% margin) — at this specific high-volume, output-heavy workload shape, Grok 4.7's cheaper output rate ($6 vs. $20 per million) makes a large practical difference to margin.

FAQ

Is Grok cheaper than ChatGPT? On the API, Grok 4.7 ($2/$6 per million tokens) is meaningfully cheaper than GPT-5.6 Sol ($4/$20) — about 2.5× cheaper overall, and more than 3× cheaper on output specifically. On consumer subscriptions, SuperGrok ($30/month) costs 50% more than ChatGPT Plus ($20/month) for a broadly similar "full access, higher limits" tier.

What is SpaceXAI? The current public brand name for the company that makes Grok, confirmed directly on its own pricing and developer-documentation pages; most existing coverage still uses "xAI."

Does SuperGrok include API access? No — the consumer subscription and the developer API are billed and provisioned entirely separately, the same pattern seen across most AI products.

Which Grok model is current — 4.6 or 4.7? Both appear on xAI's own live pages as of publication: the consumer SuperGrok plan page describes Grok 4.6, while the developer API docs page (updated September 21, 2026) lists Grok 4.7 as the new flagship. xAI's public pages do not specify whether SuperGrok subscribers already have access to Grok 4.7.

Is Grok's free tier still useful? For text chat, search and voice, yes; image and video generation were removed from the free tier during 2026 per multiple convergent sources.