Three glowing spheres in red, blue, and green with light rays converging on a black background, representing three AI chatbots being weighed against each other.
Grok, ChatGPT, and Claude aren't sold the same way, updated the same way, or moderated the same way. The pricing model is the biggest practical difference, not the model quality.

Grok, ChatGPT, and Claude all do the same basic thing: you type a question, they type back an answer. Past that, the three products come from three companies with different businesses, different money-makers, and different ideas about what an AI assistant should refuse to do. This page compares them the way someone actually shopping compares them — what does each one cost, how do you actually get it, what does “real-time information” mean for each, and what should you expect it to say no to. It closes with direct picks by situation, because that’s the actual point of a buying guide.

The quick answer

Already pay forX PremiumYou already have Grok. Don't add another subscription just to also get an AI chatbot.
→
WantThe most well-rounded optionImages, voice, video, the biggest ecosystem, most name recognition. ChatGPT Plus.
WantCareful writing, coding, long documentsThe most conservative default, no ads, no social feed attached. Claude Pro.
→
WantLive awareness of what's happening on X right nowGrok is the only one built around that as a first-class feature. SuperGrok, standalone.

The rest of this page is the reasoning behind those four picks.

Who actually makes each one

ChatGPT is OpenAI’s product. As of September 2026 the self-serve flagship family is GPT-5.6 (Sol/Terra/Luna, generally available since July 2026), with GPT-6 Astra — OpenAI’s newest and most capable model — rolling into paid ChatGPT plans from 3 September 2026 [1][2].

Claude is Anthropic’s product. The current lineup runs Haiku 4.5 (fastest/cheapest) up through Sonnet 5, Opus 5 (the default flagship), and Claude Fable 5.1, which reached general availability on 1 September 2026 as Anthropic’s most capable widely-released model [3].

Grok has the more complicated ownership story, and it matters for a buying decision because it changes who you’re actually handing your data and money to. Grok was built by xAI, which merged into SpaceX in an all-stock deal that closed 2 February 2026 — SpaceX valued at roughly $1 trillion, xAI at roughly $250 billion, a combined entity worth about $1.25 trillion [4][5]. xAI stopped existing as a separate company, and the combined AI-plus-rockets-plus-Starlink-plus-X operation rebranded as SpaceXAI in July 2026 [6]. Practically: Grok is no longer a standalone AI startup’s product, but one part of a single, very large, Musk-controlled company that also owns X (formerly Twitter) — which is why Grok can be bundled into an X subscription in the first place. The current flagship model is Grok 4.6, released 12 August 2026 with a 500,000-token context window, positioned for coding and agentic work [7][8].

What it actually costs — the real differentiator

This is the part most comparisons undersell. ChatGPT and Claude are each sold as one dedicated product with one price ladder. Grok is sold two ways at once: as its own standalone product (SuperGrok), and as a feature bundled into a social-media subscription (X Premium) that also buys you a verification checkmark, longer posts, an edit button, and reduced ads on X — none of which have anything to do with the AI model. That’s a genuinely different purchase than “I am buying an AI subscription.”

Tier bandChatGPT (OpenAI)Claude (Anthropic)Grok (xAI / SpaceXAI)
Free$0 — unlimited text chat, capped images/voice/Deep Research, ad-supported [2]$0 — full feature set (web search, artifacts, memory) at low usage limits [3]$0 — limited messages, lighter model, no image generation
~$8–10/moGo — $8/mo, same model as Free with ~10x the caps, still ad-supported [2]— no equivalent tier —X Premium — $8/mo, bundled into X’s base subscription: basic Grok access plus checkmark, reduced ads, longer posts. SuperGrok Lite — $10/mo, standalone, no X account features
~$20–40/moPlus — $20/mo, full flagship model, voice, Sora video, Deep Research, ad-free [2]Pro — $17/mo billed annually ($200/yr) or $20/mo monthly, flagship model plus Claude Code and Claude Design [3]SuperGrok — $30/mo, full Grok 4.6, DeepSearch, standalone. X Premium+ — $40/mo (~$350–395/yr), fuller Grok access bundled with ad-free X
$100+/moPro (Codex) — $100/mo, ~5x Plus’s limits, elevated coding/Codex allowance. Pro (Max) — $200/mo, ~20x Plus’s limits, effectively unmetered Sora and Advanced Voice [2]Max — from $100/mo, 5x or 20x Pro’s usage [3]SuperGrok Plus — $100/mo, higher usage across chat/image/voice. SuperGrok Heavy — $300/mo, multi-agent “Heavy” mode

Four things worth pulling out of that table. ChatGPT actually splits Pro into two prices, not one — $100/month (Pro Codex, ~5x Plus’s limits, aimed at coding workloads) and $200/month (Pro Max, ~20x Plus’s limits, the “barely any ceiling” tier); OpenAI added the $100 rung in April 2026, so treat it as current rather than a stale or regional figure [2]. Grok’s cheapest paid path is not an AI purchase — $8/month gets you basic Grok riding along on a social subscription, and the real apples-to-apples comparison is ChatGPT Go at the same $8, a dedicated AI tier with no social platform attached. At the $100/mo rung specifically, SuperGrok Plus, ChatGPT Pro (Codex), and Claude Max (5x) are directly cost-comparable, so pick by which model you already prefer, not price — and remember ChatGPT and Claude both also have a second, pricier rung ($200) above that one. And xAI stopped counting individual messages in June 2026, switching to a shared weekly compute pool across chat, image, voice, and API, shown as a percentage in account settings — so a “10 messages per 2 hours”-style limit no longer describes any current Grok tier.

Every figure above changes without much notice on all three sides; treat this table as a September 2026 snapshot. See ChatGPT Free vs Plus vs Pro for OpenAI’s full ladder, including the Business and Enterprise tiers this page skips.

Real-time information: all three have it now, but not the same way

Grok’s headline pitch has always been that it can see what’s happening on X right now — trending topics, a specific account’s recent posts, reaction to something that happened an hour ago. That’s a genuine structural advantage: Grok’s parent company owns the platform the data lives on, rather than scraping a public API the way a third party would. If your question is “what’s the current reaction to [event] on X,” Grok is the only one of the three built around answering that as a first-class feature.

That said, this is no longer Grok’s exclusive lane. ChatGPT has had built-in web browsing for a while, and Claude added web search across every tier, including Free [3]. For a general “what’s the latest on X topic” question answered from the open web, all three can now do it — the real differentiator is specifically live, native access to X’s own post stream, not internet access in general.

Content policy: the part worth being genuinely careful about

xAI has consistently marketed Grok around fewer refusals and a willingness to engage with topics other assistants decline — Grok’s own published system-prompt history has, at various points, instructed the model not to shy away from “politically incorrect” claims when well-substantiated [9]. That positioning is real and it’s a legitimate reason some users prefer Grok for political commentary, satire, or blunt takes other assistants soften.

It’s also directly connected to Grok’s worst public incident. In December 2025 and January 2026, Grok Imagine’s “Spicy” image-generation mode was shown capable of producing sexualized and deepfake imagery of real people, including content resembling minors. Indonesia and Malaysia blocked Grok over it in mid-January 2026 — Malaysia lifted its block on 23 January after xAI added safeguards, but kept the product under ongoing monitoring [10][11]. The European Commission opened a formal Digital Services Act investigation into Grok and X’s risk-management obligations in January 2026, ordering X to retain all related internal data through the end of 2026; that investigation was still open, with no fine specific to Grok announced, as of this writing [12][13]. UK and French regulators opened parallel scrutiny. xAI responded by tightening Grok Imagine’s moderation substantially from mid-January 2026 onward.

Neither ChatGPT nor Claude has faced a comparable regulatory action over generated content as of this writing. That’s not the same as saying either is free of controversy in other respects — copyright litigation and labor disputes touch every major AI lab — but specifically on the axis this section is about, permissiveness in what the model will generate, Grok is the one of the three with an open regulatory investigation attached to it. If you’re buying this for a shared family device, a workplace-monitored account, or anyone else you’d rather not have exposed to an unsettled situation, that’s a genuinely relevant fact, not a reason to write Grok off outright for every use case.

General capability: close enough that your own workload should decide

All three are frontier-competitive as of September 2026, and vendor benchmark claims from any of the three should be tested against your actual use case rather than trusted at face value. What holds up as a reasonable generalization:

  • ChatGPT has the broadest built-in feature set for a single subscription — native image generation, Sora video, voice mode, the largest third-party plugin/app ecosystem, and by a wide margin the largest user base of the three. The tradeoff is at the bottom of its ladder: Free and Go are the only ad-supported tiers of the three products compared here.
  • Claude is consistently well-regarded for coding, long-document analysis, and careful instruction-following, with a deliberately narrower product surface — no native image, video, or voice generation — and the most conservative default safety posture of the three.
  • Grok 4.5 and 4.6 are marketed heavily around coding and agentic-task benchmarks at lower per-token API pricing than Claude’s or GPT’s top tiers, with xAI positioning Grok 4.5 as competitive with Anthropic’s Opus-class models specifically [7].

None of the three has a categorical, durable lead across every task. Use the free tier of each on your actual work before paying for any of them.

Who should actually pick what

The quick answer at the top covers the four most common cases. A few more specific ones:

You want the cheapest possible entry into a real, dedicated AI subscription, not a social-media add-on. ChatGPT Go at $8/month is the actual comparison point to X Premium’s $8 — same price, a genuinely different kind of purchase.

You’re specifically weighing a $100+/month “unlimited-feeling” plan. SuperGrok Plus, ChatGPT Pro (Codex), and Claude Max (5x) all sit at $100/month; ChatGPT Pro (Max) and Claude Max (20x) both step up to $200 for a materially higher ceiling. Pick by which model you already preferred on the free tier — price stops being the deciding factor up here.

You’re buying for a shared, family, or workplace-monitored account. Weight the content-policy section above more heavily than the capability section: Claude’s conservative default and ChatGPT’s conventional moderation both have longer, less eventful track records on this specific axis than Grok’s does right now.

You’re not sure yet. All three have a genuine free tier. Run your actual, real questions through each one before paying for any of them — a benchmark chart is a worse predictor of what you’ll like than five minutes of your own use.

What this guide can’t tell you

  • Exact current prices, especially for Grok — xAI’s and X’s own pricing pages block automated access, so the Grok figures in the table above are cross-checked against several independent trackers rather than fetched directly; confirm on grok.com and x.com/i/premium before buying.
  • How the EU’s Grok investigation resolves. It was open, with no fine or ruling specific to Grok, as of publication. A future outcome could change Grok’s feature set or availability in the EU specifically.
  • Regional pricing. Figures here are US web pricing; app-store pricing (Apple/Google take a cut) and non-US regional pricing both commonly differ from the numbers above.
  • Which one “feels” best for your writing or workflow. That’s genuinely subjective. Try the free tier of each before paying for any of them.

Further reading

Sources

  1. This wiki, OpenAI API : GPT-5.6 and GPT-6 Astra model lineup and rollout timeline, last updated 3 September 2026.
  2. This wiki, ChatGPT Free vs Plus vs Pro vs Team/Business/Enterprise (4 September 2026), itself sourced to OpenAI’s own pricing and help-center pages, fetched September 2026.
  3. Anthropic, Claude pricing page, fetched 4 September 2026: https://claude.com/pricing
  4. TechCrunch, “Elon Musk’s SpaceX acquires xAI” (2 February 2026): https://techcrunch.com/2026/02/02/elon-musk-spacex-acquires-xai-data-centers-space-merger/
  5. CNN, “Elon Musk’s SpaceX acquires xAI, merging his two most ambitious companies” (2 February 2026): https://www.cnn.com/2026/02/02/tech/spacex-acquires-xai-elon-musk
  6. Engadget, “xAI is now officially known as SpaceXAI” (July 2026): https://www.engadget.com/2209300/xai-now-officially-spacexai/
  7. xAI, “Introducing Grok 4.5” (July 2026): https://x.ai/news/grok-4-5 ; TechCrunch, “SpaceXAI releases Grok 4.5, which Elon describes as an ‘Opus-class’ model” (8 July 2026): https://techcrunch.com/2026/07/08/spacexai-releases-grok-4-5-which-elon-describes-as-an-opus-class-model/
  8. xAI/SpaceXAI, Grok developer documentation — models and pricing, fetched 4 September 2026: https://docs.x.ai/developers/models , https://docs.x.ai/developers/pricing
  9. xAI, published Grok system-prompt history: https://github.com/xai-org/grok-prompts
  10. PBS NewsHour, “Malaysia and Indonesia become the first countries to block Musk’s chatbot Grok over sexualized AI images” (January 2026): https://www.pbs.org/newshour/world/malaysia-and-indonesia-become-the-first-countries-to-block-musks-chatbot-grok-over-sexualized-ai-images
  11. TechCrunch, “Indonesia blocks Grok over non-consensual, sexualized deepfakes” (11 January 2026): https://techcrunch.com/2026/01/11/indonesia-blocks-grok-over-non-consensual-sexualized-deepfakes/
  12. European Commission, “Commission investigates Grok and X’s recommender systems under the Digital Services Act”: https://digital-strategy.ec.europa.eu/en/news/commission-investigates-grok-and-xs-recommender-systems-under-digital-services-act
  13. France24, “EU orders Musk’s Grok AI to keep data after nudes outcry” (8 January 2026): https://www.france24.com/en/live-news/20260108-eu-orders-musk-s-grok-ai-to-keep-data-after-nudes-outcry
  14. This wiki, xAI Grok , last updated 29 June 2026 — predates the SpaceXAI rebrand and should not be relied on for current company or ownership framing; the SpaceX-merger sources above (4–6) are more current.
  15. Cross-checked Grok/X Premium consumer pricing trackers (ai-toolbox.co, costbench.com, geotoolbox.ai, suprmind.ai), fetched 4 September 2026 — xAI’s and X’s own consumer pricing pages returned access-blocked responses to direct automated fetch during research for this article, so the Grok consumer-tier figures above are corroborated secondary reporting, not directly primary-sourced the way the API pricing and Claude figures are. Confirm current prices at grok.com and x.com/i/premium before purchasing.