Updated 30 October 2026: Grok is now sold through The X Subscription, which replaced the SuperGrok plans for new customers on 12 October and bundles Grok, Grok Bot, Cursor and X. ChatGPT’s paid plans moved to GPT-6 in normal chats in October, and OpenAI added GPT-6.1 Terra on 26 October. We have rewritten the plans section and updated the model table, voice, moderation and FAQ. Our verdict is unchanged.
Grok is xAI’s chatbot, built into X and sold on its own at grok.com. xAI has since merged into SpaceX and now signs its pages as SpaceXSI, but Grok keeps its name. ChatGPT is OpenAI’s, and it’s still the one most people mean when they say “AI chatbot”. As of September 2026 the two models are closer in quality than their reputations suggest. The differences lie in everything around the model. Grok reads X and ChatGPT doesn’t. Grok makes video in the chat, while ChatGPT makes images. The plans are priced differently, and each company has had a very different kind of safety problem in public. We put this together from both companies’ product pages, developer documentation and model cards, first checked on 28 and 29 September 2026 and rechecked on 30 October 2026. We haven’t run our own head-to-head tests.
- Grok’s edge is live data and price. Every Grok plan, including the free one, searches the web and X as things happen. On the API, Grok 4.7 costs $2 in and $6 out per million tokens; GPT-6 Astra costs $10 and $50.
- Grok now comes in one bundle. Since 12 October, The X Subscription covers Grok, Grok Bot, Cursor and X. It starts at $8 for Plus, and Grok Bot and video need Premium at $30. ChatGPT Plus is $20.
- ChatGPT’s edge is the top model and the agent tooling. GPT-6 Astra, out since 3 September 2026, is OpenAI’s most capable model and the one it recommends for computer use, coding and professional documents. Grok has nothing like ChatGPT’s Codex, Sites or built-in browser.
- Grok makes images and video in the app; ChatGPT makes images. Grok’s free tier has limited image generation, and video starts with Premium. ChatGPT’s free tier also includes some image generation.
- Both have a safety record, and they’re different ones. Ofcom and the UK Information Commissioner opened investigations into X and xAI in January and February 2026 over sexualised deepfakes made with Grok. OpenAI says GPT-6 Astra crosses its own “Critical” threshold for cyber capability, and ships it with refusals and monitoring for that reason.
Which model you actually get
Start with the names, because both companies changed them over the summer. xAI’s newest model is Grok 4.7, announced on 21 September 2026 as “our most capable model for coding and knowledge work”. It launched through the xAI API, Grok Build (xAI’s coding agent) and Cursor, and the model card said the consumer apps would get it “at a later date”. xAI’s consumer pages no longer name a version: the pricing page just promises the “Latest Grok models” from Premium up. Independent reports say Grok 4.7 has run Expert mode on grok.com since 3 October. xAI hasn’t announced that itself, so treat it as likely rather than confirmed. Grok 4.6, from 14 August, is still listed on the API.
OpenAI’s line-up is the GPT-6 family. GPT-6 Astra arrived on 3 September 2026, billed as “the world’s most intelligent and aligned model”, and on 22 September OpenAI added GPT-6 Sol and GPT-6 Luna, two cheaper models for coding and high-volume work. On 29 September it followed with GPT-6.1 Sol, an upgrade to Sol that OpenAI says “nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work” at a fifth of Astra’s token prices; it’s in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, but not yet in Chat. On 7 October GPT-6 reached the normal Chat tab for paid plans. OpenAI’s model help page, updated at the end of October, says GPT-6 Sol now powers Instant through Extra High on Plus, Pro, Business and Enterprise, and the Pro option runs GPT-6 Pro on GPT-6 Astra. Free and Go users still get GPT-5.6 Luna, even though OpenAI’s 7 October release note had said GPT-6 Luna was coming to them. On 26 October OpenAI added GPT-6.1 Terra, a cheaper model aimed at computer use, in ChatGPT Work and Codex.
| Model | Where you get it | API price per 1M tokens (in / out) | Context | Knowledge cutoff |
|---|---|---|---|---|
 Grok 4.7grok-4.7 · 21 Sep 2026 | API, Grok Build, Cursor; reportedly Expert mode on grok.com | $2 / $6 (under 200K prompt tokens; $4 / $12 above) | 500K | Jun 2026 (pretraining) |
| Grok 4.6grok-4.6 · 14 Aug 2026 | API | $2 / $6 | 500K | Not stated on docs page |
 GPT-6 Astragpt-6-astra · 3 Sep 2026 | ChatGPT Plus, Pro, Business, Enterprise; API | $10 / $50 | 1.05M | 30 Apr 2026 |
| GPT-6 Solgpt-6-sol · 22 Sep 2026 | ChatGPT Chat on Plus and up (since 7 Oct), Codex, Work, API | $2 / $10 | 1.05M | 20 Apr 2026 |
| GPT-6.1 Solgpt-6.1-sol · 29 Sep 2026 | Codex, ChatGPT Work, API; not yet in Chat | $2 / $10 | Not yet confirmed | Not yet confirmed |
| GPT-6 Lunagpt-6-luna · 22 Sep 2026 | Codex, ChatGPT Work, API | $0.10 / $0.50 | 1.05M | 18 May 2026 |
| GPT-6.1 Terragpt-6.1-terra · 26 Oct 2026 | Codex, ChatGPT Work, API; not in Chat | $1 / $5 | Not yet confirmed | Not yet confirmed |
Two things in that table settle a lot of arguments. Grok 4.7, GPT-6 Sol and GPT-6.1 Sol cost the same on input, $2 per million tokens, but Grok is cheaper on output, $6 against $10, and all three are a fraction of Astra. The context windows aren’t equal either. GPT-6 models take just over a million tokens, Grok 4.7 takes 500,000, and xAI doubles its price once a prompt goes past 200,000 tokens. In ChatGPT itself, reasoning models get a 256,000-token window on Plus and 400,000 on Pro. xAI doesn’t publish a per-plan figure for grok.com.
What Grok does differently
The big one is X. Grok “searches the web and X live”, as xAI’s product page puts it, and the pitch is “breaking news, trends, and posts as they happen”. No other mainstream chatbot has that feed, because no other company owns it. ChatGPT searches the web, and does it well, but it sees X the way the rest of the internet does, which is barely. If you ask what people are saying right now about a live event, a launch or a public row, Grok has a source ChatGPT simply doesn’t.
Then there’s generation inside the chat itself. Grok Imagine makes images and video from a text prompt or a reference photo, and edits them with follow-up prompts, all in the same thread. xAI’s product page promises images “up to 2K resolution” and text-to-video “up to 15 seconds at 720p”, and the pricing page puts video in the Premium plan and above. ChatGPT generates images, and its Pro plan promises “unlimited and faster image creation”, but video generation isn’t listed as a feature anywhere on the ChatGPT pricing page.
xAI also has a multi-agent mode, where “multiple agents work in parallel for deeper answers on the hardest questions” and each one shows its reasoning. On top of that there’s Grok Bot, xAI’s agents that work on their own cloud computer, sign in to your apps and keep working when your laptop is closed. It comes with The X Subscription from Premium up, and with older SuperGrok, Cursor and X Premium+ plans.
Voice is where Grok made its name in 2025, with personalities that ran from a meditation guide to an “unhinged” option. xAI’s current product page describes voice simply as “natural back-and-forth with sub-second latency”, and the unhinged personality doesn’t appear on any of the official pages we checked this week, so don’t count on it. What xAI does document is a speech-to-speech model on the API. Grok Voice 3.0 replaced Voice 2.0 in October at the same $0.08 a minute.
What ChatGPT does that Grok does not
ChatGPT’s answer to all that is its top model and the tools around it. OpenAI calls GPT-6 Astra “the world’s best computer use model” and spells out what that means: filling in online forms, updating records in a CRM, running front-end QA on a website it has just built, installing and testing software. Developers can build agents like that on OpenAI’s, Google’s and Anthropic’s APIs, and we compare the three computer use tools and their costs separately. On OpenAI’s own numbers Astra scores 57.9 percent on Terminal-Bench 4.0, a benchmark of long terminal tasks, where xAI reports 37.6 percent for Grok 4.7. Those are the companies’ figures, not ours, and they come from two different launch posts, but a gap that size is hard to wave away.
Around the model, ChatGPT has more places to put an agent to work: Codex for coding, ChatGPT Work on the desktop, a built-in browser, Sites for hosting web apps from a prompt, scheduled tasks, projects and shared projects. xAI has Grok Build, its own open-source coding agent, add-ins for Word, Excel, PowerPoint and Outlook, and Grok Bot. OpenAI’s answer to Grok Bot arrived on 29 September: dots, always-on agents powered by GPT-6 Astra that have their own cloud computer and connect to more than 4,000 apps. They are rolling out gradually to eligible Pro and Business Premium users, but Pro access excludes the European Economic Area, Switzerland and the UK at launch. The two are converging, but this month OpenAI’s pricing page lists more of these features, and on more plans.
Plans and prices
Both are free to try and both charge for more. What differs is what the free tier leaves out. Grok’s free plan gives limited chat and limited image generation, with real-time web and X search; video comes with Premium. ChatGPT’s free plan runs GPT-5.6 Luna, with “limited and slower image generation” and limited voice and deep research.
p>The bigger change is on xAI’s side. Since 12 October, new customers no longer buy SuperGrok. They buy The X Subscription, one plan with a single weekly usage pool shared across Grok, Grok Bot and the Cursor coding editor, plus X features such as the checkmark and fewer ads. Existing SuperGrok, Cursor and X Premium subscribers keep their plans. Prices exclude VAT.
On the API the gap is wider than in the apps. A workload that runs on GPT-6 Astra at $10 and $50 would run on Grok 4.7 at $2 and $6, roughly a fifth to an eighth of the cost per token. Against GPT-6 Sol and the new GPT-6.1 Sol it’s closer, $2 and $10 versus $2 and $6, and against GPT-6 Luna at $0.10 and $0.50 Grok has nothing to offer in its current line-up. Both companies sell a “fast” variant at double the price for double the speed. OpenAI charges $10 per thousand web-search calls on top of tokens; xAI prices image, video and voice generation as separate line items. We cover another low-cost rival in DeepSeek vs ChatGPT.
Moderation and safety
This is where the two differ most, and where it’s easiest to write nonsense, so we’re sticking to what regulators and the companies themselves have published. In January 2026 Ofcom, the UK communications regulator, contacted X about “concerning reports of the Grok AI chatbot account on X being used to create and share demeaning sexual deepfakes of real people, including children”. It opened a formal investigation into X under the Online Safety Act on 12 January. On 3 February the Information Commissioner’s Office opened its own investigation into X Internet Unlimited Company and X.AI LLC, covering “both the development and deployment of Grok” and whether “appropriate safeguards were implemented”. Ofcom’s 3 February update notes that X “has said it has implemented measures to try and address the issue”, that the European Commission had opened an investigation of its own on 26 January, and that Ofcom’s case remains open.
Two details from that update matter if you’re choosing between these tools. Ofcom said it was “not investigating xAI at this time” because the standalone Grok app, where a user chats one-to-one with the model, largely falls outside the Online Safety Act; only X, the social network, is in scope. Image generation on Grok is now heavily restricted for free users. xAI’s pricing page lists it as “Limited” on the free plan, and video only comes with Premium.
p>Ofcom gave another update on 14 October 2026, as reported by Reuters. Its investigation has “moved into a new phase” focused on X’s handling of illegal content. UK users are now blocked from viewing, generating or sharing pornographic content through Grok, and companion characters can’t hold sexually explicit chats. There is still no decision; Ofcom will issue a provisional view first. The ICO investigation is also still open.
xAI sets out its own position in the Grok 4.7 model card, dated 21 September 2026. It describes “an entirely new safeguard stack”, reports 0.0 percent compliance on its internal child-safety evaluation across Grok 4.5, 4.6 and 4.7, and says that on consumer surfaces the model is “additionally augmented with runtime input and topical filters” for CSAM, self-harm and weapons. It also says the goal is “to preserve legitimate uses” and to avoid “over-refusal on benign or hypothetical discussions”. Grok has always been positioned that way, as deliberately more permissive than its rivals.
OpenAI’s safety story this year is about capability rather than content. Its GPT-6 Astra announcement says the model “meets the Critical threshold in cybersecurity under our Preparedness Framework”, that it “will refuse to comply with more advanced cybersecurity tasks such as creating proof-of-concept exploits”, and that OpenAI is “deploying misalignment monitoring in production” that can pause or stop a task. The same post mentions “the Hugging Face incident” and says OpenAI built an evaluation from it in which GPT-5.6 Sol, without production safeguards, “went beyond the authorized target 48% of the time”, while Astra did so in 0 percent of cases. OpenAI doesn’t describe the incident any further there, so neither will we. Both companies now publish numbers on their own failure modes. They just aren’t the same failure modes.
Which one to pick
+Reasons to pick Grok
- You want live X and web data in every answer, on the free plan
- You make images and short video inside the chat, up to 15-second clips at 720p
- You pay per token: $2 / $6 for Grok 4.7 against $10 / $50 for Astra
- You’d also use Grok Bot and Cursor, which The X Subscription bundles from $30
+Reasons to pick ChatGPT
- You want the strongest model on computer use and long agent tasks
- You need a million-token context, or 400K in the app on Pro
- You use Codex, Sites, a built-in browser or shared projects
- You want stricter refusals by default, and a free tier with image generation
For most people ChatGPT is still the safer default. The top model is better at the tasks that matter to agents, the free tier does more, and the plans are easier to understand. Grok suits a particular kind of user. You live on X, you want a chatbot that reads it, you want video without opening another tool, and you don’t mind that the company’s moderation is still under investigation in the UK and the EU. On the API we’d lean the other way. Grok 4.7 at $2 and $6 is one of the cheapest frontier-class models you can buy, and if your workload doesn’t need Astra’s computer use or a million tokens of context, give it a proper trial. If neither suits you, we look at nine other assistants in our guide to ChatGPT alternatives.
Frequently asked questions
Is Grok better than ChatGPT?
Not overall. On the companies’ own benchmarks GPT-6 Astra leads on agent and terminal tasks, while Grok 4.7 costs far less and can read X live. Pick Grok for real-time social data, in-chat video and API cost; pick ChatGPT for computer use, long context and agent tooling.
Is Grok free?
Yes, with limits. The free plan includes limited chat and image generation with real-time web and X search. Paid plans are part of The X Subscription: Plus $8, Premium $30 (adds video and Grok Bot), Super $100 and Ultra $200 a month, checked on x.ai on 30 October 2026.
Which Grok model do I get on grok.com?
xAI no longer names it on its consumer pages. Grok 4.7, released on 21 September, went to the API, Grok Build and Cursor first, and independent reports say it has run Expert mode on grok.com since 3 October. Premium and above get the “Latest Grok models”.
Does ChatGPT have access to X?
No. ChatGPT searches the open web and can browse pages, but it has no feed of X posts. Grok’s “real-time X integration” is unique to xAI because X and xAI are under common ownership.
How much cheaper is Grok on the API?
Grok 4.7 is $2 per million input tokens and $6 per million output tokens for prompts under 200,000 tokens. GPT-6 Astra is $10 and $50; GPT-6 Sol and GPT-6.1 Sol are $2 and $10; GPT-6.1 Terra is $1 and $5; GPT-6 Luna is $0.10 and $0.50. Grok doubles its rate above 200,000 prompt tokens and caps context at 500,000, where GPT-6 models take about 1.05 million.
Sources and method
Model names, prices, context windows and plan contents were taken from xAI’s and OpenAI’s own pages on 28 and 29 September 2026 and rechecked on 30 October 2026, when we also read OpenAI’s ChatGPT release notes and xAI’s X Subscription pricing page in a browser. The claim that Grok 4.7 runs Expert mode on grok.com comes from independent reports, not from xAI. Benchmark figures are the companies’ published claims from their launch posts and model cards, not our measurements; the two companies report on different benchmark sets and we only compare where both publish the same test. Statements about investigations are quoted from Ofcom and the Information Commissioner’s Office. ChatGPT Go’s price was visible to us only in euros on the localised pricing page; Plus and Pro dollar prices are from OpenAI’s help centre.
- xAI, Introducing Grok 4.7 (release date, API price, benchmark and safeguard claims)
- xAI, Grok 4.7 model card (availability by surface, data cutoff, jailbreak and child-safety evaluations, runtime filters)
- xAI, Models and pricing (token, image, video and voice API prices; context windows)
- xAI, Grok product page (real-time web and X search, Imagine, multi-agent mode, voice)
- xAI, Pricing (The X Subscription tiers, prices and features, checked 30 October 2026)
- xAI, Introducing The X Subscription (12 October 2026; what each tier replaces)
- xAI Docs, Release notes (Grok Voice 3.0, Imagine API changes)
- OpenAI, GPT-6 Astra: A new generation of intelligence (release, benchmarks, cyber threshold, monitoring, Hugging Face evaluation)
- OpenAI, Models and Pricing (GPT-6 Astra, Sol and Luna prices, context, cutoffs)
- ChatGPT, Pricing (plan contents, per-plan context windows)
- OpenAI Help Center, What is ChatGPT Plus? (Plus $20, Pro $100 and $200)
- Ofcom, Investigation into X, and scope of the Online Safety Act (3 February 2026 update)
- ICO, ICO announces investigation into Grok (3 February 2026)
- OpenAI Help Center, ChatGPT release notes (GPT-6 in Chat from 7 October, GPT-6.1 Terra on 26 October, Pro allowances on 29 October)
- OpenAI Help Center, GPT-6 and other models in ChatGPT (which model each plan runs in Chat, checked 30 October 2026)
- Reuters via TimesLIVE, Ofcom update on X (14 October 2026)
- OpenAI, Introducing GPT-6.1 Sol (release, availability, API price)
- OpenAI Help Center, About ChatGPT Pro tiers (Pro 100, Pro 200 and Pro 500 prices, Pro 200 reopening and allowance, Ultrafast)
- OpenAI, DevDay 2026 Recap (Pro 500 allowance)
- OpenAI, Introducing dots (what dots are, who gets them, excluded regions)
