Published Updated
In one or two sentences, before anything else.
Grok is the chatbot from SpaceXSI, the Musk company formerly called xAI and then SpaceXAI, sold since October 13, 2026 in the X Subscription. Since January 2026 it has faced country blocks, investigations and lawsuits over sexualized deepfakes, and Canada's Privacy Commissioner found in June 2026 that X Corp. and xAI broke federal privacy law.
What it can do, one sentence each, and whether we actually tested it.
Answers with real-time access to posts on X, which Zapier says other chatbots cannot match for breaking news.
Zapier's roundup, last updated November 2025.
Lets you pick between fast replies and heavier reasoning; Artificial Analysis lists Grok 4.7 at low, high and xhigh.
Makes images and short videos from prompts. Engadget lists image and video creation in the 30 USD Premium tier.
Regulators in several countries acted over sexualized deepfakes made with Grok.
Spoken conversation with Grok, listed in the 8 USD Plus tier.
Per Engadget, October 13, 2026.
A command-line coding agent included from the Plus tier.
In July 2026 it was found uploading whole repositories, including secrets (The Register, TNW).
An agent that carries out tasks, included from the Premium tier.
Per Engadget and TechCrunch, October 13, 2026.
Grok 4.7 costs 2 USD per million input tokens and 6 USD per million output tokens for prompts under 200,000 tokens, with a 500,000-token context window.
Per heise, September 22, 2026.
The people it helps and the problems it solves, and where it ranks.
Built for
Problems it solves
Graded in
Public tests other people ran: what they gave it and how it did.
Case 1
Input
The full Intelligence Index task set at the xHigh reasoning level, run on Grok 4.7, GPT-6 Astra and other models.
What happened
Grok 4.7 scored 46, two points more than Grok 4.6. It wrote about 81,000 output tokens per task against 27,000 for GPT-6 Astra Max, so it cost 3.74 USD a task against 3.26 USD despite a lower price per token.
Case 2
Input
116 private tasks answered without web search or code execution, scored for accuracy.
What happened
Grok 4.5 scored 83.5%, sixth, behind Claude 5 Fable with thinking at 91.4% and GPT-5.5 at 88.8%. It cost 1.0 USD for the whole set against 10.4 USD for Claude 5 Fable with thinking. The test covers Grok 4.5, not the newer 4.7.
Case 3
Input
Hundreds of public Grok Imagine links, plus two of the prompts behind them re-run on ChatGPT, Meta AI, Claude and Gemini.
What happened
Wired found dozens of sexualized images and videos, some depicting celebrities and a US politician without consent. ChatGPT, Meta AI and Claude refused both prompts; Gemini made one image and refused the other. The content became unavailable after Wired contacted the company.
Case 4
Input
A harmless one-word prompt in a code repository, with instructions not to open any files, while network traffic was recorded.
What happened
Grok Build still uploaded the whole repository with its Git history, including secrets deleted months earlier, to a cloud storage bucket. Uploads stopped after a server-side change, and Musk promised to delete earlier uploads; The Register could not verify the deletion.
What it does badly, with the evidence, and what we could not settle.
Weak
Image safeguards have failed before, and regulators are not convinced they work now.
Evidence: AI Forensics found about 2% of 20,000 Grok images appeared to show people 18 or younger (AP, January 6, 2026). Canada's Privacy Commissioner found on June 11, 2026 that X Corp. and xAI broke federal privacy law, and Wired reports the finding says the safeguards have not been shown to work.
Looser content rules by design.
Evidence: Wired reports xAI's terms say Grok may respond with "sexual situations" and that it has offered "Spicy" and "Unhinged" modes.
More expensive per task than its token price suggests.
Evidence: Artificial Analysis puts Grok 4.7 at 3.74 USD a task at xhigh against 3.26 USD for GPT-6 Astra Max, because it writes about three times as many tokens (heise).
Behind the leading models on independent tests.
Evidence: Artificial Analysis scores Grok 4.7 at 46 against 58 for Claude Opus 5.5 and 53 for GPT-6 Astra and Gemini 4 Argon.
Its coding agent sent whole repositories to the cloud.
Evidence: In July 2026 Grok Build uploaded full repositories, including old secrets, even when told not to open files (The Register, TNW).
Who should not use it now: children and teens, schools and workplaces that need strict filters, and anyone editing photos of real people.
Evidence: Wired still found sexualized deepfakes hosted on grok.com in June 2026, five months after the restrictions, and Canada's Privacy Commissioner said the safeguards had not been shown to work.
Uncertain
How many images a free account can make today.
Why unsure: AlloyPress reported free users lost image and video generation on March 19, 2026, with no official statement; Engadget lists image and video creation under Premium. We could not read grok.com/plans.
Which Grok model the apps run now.
Why unsure: heise reported on September 22, 2026 that Grok 4.7 would reach grok.com, the apps and X later; we found no confirmation that it has.
The outcome of the investigations.
Why unsure: Ofcom opened a formal investigation on January 12, 2026 and can fine up to 10% of qualifying worldwide revenue (RTÉ); we found no decision yet.
Whether earlier Grok Build uploads were deleted.
Why unsure: Musk promised deletion, but The Register and TNW could not verify it.
The numbers, each with its source and the date we checked it.
| Spec | Value | Note and source |
|---|---|---|
| Maker | SpaceXSI (SpaceX Super Intelligence), part of SpaceX | Source: techcrunch.comchecked |
| Former names | xAI until 2026, then SpaceXAI; Musk announced the rename to SpaceXSI in October 2026 | Source: engadget.comchecked |
| Ownership | SpaceX acquired xAI on February 2, 2026 | Source: finance.yahoo.comchecked |
| Plans (X Subscription, from October 13, 2026) | Free; Plus 8 USD, Premium 30 USD, Super 100 USD, Ultra 200 USD a month | Source: techcrunch.comchecked |
| Free tier | Limited chat and agent usage | Source: techcrunch.comchecked |
| Image and video creation | Listed in Premium, 30 USD a month | Source: engadget.comchecked |
| Grok 4.7 API price, prompts under 200,000 tokens | 2 USD input, 6 USD output per million tokens; 500,000-token context | Source: heise.dechecked |
| Artificial Analysis Intelligence Index | Grok 4.7 (xhigh) 46; Claude Opus 5.5 58; GPT-6 Astra 53; Gemini 4 Argon 53 | Source: artificialanalysis.aichecked |
| Temporary country blocks | Indonesia and Malaysia in January 2026, lifted by February 1, 2026 | Source: straitstimes.comchecked |
| Canada privacy finding | June 11, 2026: X Corp. and xAI broke federal private-sector privacy law | Source: priv.gc.cachecked |
| Website visits, September 2026 (grok.com) | 211 million (−3.2% on August); 5th of 20 on AICPB's AI chatbots list; 9th of 20 on AICPB's AI tools worldwide list | Source: mp.weixin.qq.comchecked |
Answer first, then why, an example, the limits and what to do next.
Indonesia and Malaysia lifted their temporary Grok blocks in early 2026. Malaysia lifted its block on January 23 after X confirmed new controls, and Indonesia let Grok resume on February 1 after X wrote to the government; both say they keep monitoring it.
Limits
Bloomberg reported the Philippines also blocked Grok in January; we could not open a report on when that block ended.
Grok was used from late December 2025 to make sexualized images of real people without consent, including some who appeared to be minors. AI Forensics found about 2% of 20,000 images appeared to show people 18 or younger. Ofcom opened an investigation, Indonesia and Malaysia blocked Grok for a time (Bloomberg reported the Philippines did too), and Canada's Privacy Commissioner found in June 2026 that X Corp. and xAI broke privacy law.
Limits
Wired found sexualized deepfakes still hosted on grok.com in June 2026, though researchers say such images have become harder to make.
Grok has a free tier with limited chat and agent usage, TechCrunch reported on October 13, 2026. Paid tiers cost 8, 30, 100 and 200 USD a month; Engadget lists image and video creation and the latest Grok models from the 30 USD Premium tier.
Limits
We could not read grok.com/plans, so the free image allowance is unconfirmed.
xAI now goes by SpaceXSI, its second new name in 2026. SpaceX acquired xAI on February 2, 2026; by May the company called itself SpaceXAI and Musk said xAI would be dissolved as a separate company; in October Musk announced the rename to SpaceXSI, short for SpaceX Super Intelligence. Grok kept its name.
Grok is not a good choice for children or teenagers. Wired reports xAI's terms say Grok may respond with sexual situations, it has offered Spicy and Unhinged modes, and Canada's Privacy Commissioner said its image safeguards had not been shown to work.
Grok trails ChatGPT, Claude and Gemini on the Artificial Analysis leaderboard, and its main edge is live content from X. Grok 4.7 scores 46 against 53 to 58 for the others' top models. In Wired's June 2026 test, ChatGPT, Meta AI and Claude refused prompts that had been used on Grok.
Why
Kagi's offline benchmark is kinder: Grok 4.5 placed sixth at 83.5%, ahead of Gemini 3 Pro, at low cost.
This page is built from what can be checked on October 17, 2026: regulators' statements, news reports from TechCrunch, Engadget, The Verge, Reuters, AP, Wired, RTÉ, The Straits Times, The Register, TNW and heise, and public benchmarks from Artificial Analysis and Kagi. Every fact names who said it. We describe the deepfake scandal only as far as needed to judge the product.
According to AICPB's traffic rankings, grok.com had 211 million visits in September 2026, down 3.2% on August: 5th among AI chatbots and 9th among AI tools worldwide.
| Date | Event | Source |
|---|---|---|
| Dec 25, 2025 to Jan 1, 2026 | AI Forensics reviews 20,000 Grok images and finds about 2% appear to show people 18 or younger | AP via Fortune |
| Early January 2026 | xAI restricts image generation for most users on X | Bloomberg via The Straits Times |
| January 2026 | Indonesia and Malaysia temporarily block Grok; Bloomberg reports the Philippines did too | RTÉ, The Straits Times |
| January 12, 2026 | Ofcom opens a formal investigation into X under the UK Online Safety Act | RTÉ |
| January 23, 2026 | Malaysia lifts its block after X confirms new controls | The Straits Times |
| February 1, 2026 | Indonesia lets Grok resume | The Straits Times |
| February 2, 2026 | SpaceX acquires xAI | Reuters |
| May 6, 2026 | The company calls itself SpaceXAI; Musk says xAI will be dissolved as a separate company | The Verge |
| June 11, 2026 | Canada's Privacy Commissioner finds X Corp. and xAI broke federal privacy law; Wired finds sexualized deepfakes still hosted on grok.com | OPC, Wired |
| July 12 to 14, 2026 | A researcher finds the Grok Build coding tool uploading whole code repositories | The Register, TNW |
| September 22, 2026 | Grok 4.7 opens to developers | heise |
| October 2026 | Musk announces a second rename, to SpaceXSI; Grok joins a single X Subscription | Engadget, TechCrunch |
So the maker has been renamed twice in 2026: xAI became SpaceXAI, which is becoming SpaceXSI (SpaceX Super Intelligence). The product is still called Grok.
Since October 13, 2026, Grok is sold inside the X Subscription with X Premium and the Cursor coding tool.
| Tier | Price a month | Grok features named in reports |
|---|---|---|
| Free | 0 | Limited chat and agent usage |
| Plus | 8 USD | Grok Chat, Voice, Imagine, Build and the API |
| Premium | 30 USD | Adds Grok Bot, the latest models, AI images and video, faster replies |
| Super | 100 USD | Priority access to models at peak times |
| Ultra | 200 USD | Earliest access and 20 times the usage of Plus |
Sources: Engadget and TechCrunch. We could not read grok.com/plans, so the exact free image allowance is unconfirmed.
| Grok | ChatGPT, Claude, Gemini | |
|---|---|---|
| Artificial Analysis Intelligence Index | 46 (Grok 4.7) | 53 to 58 for their top models |
| Live social media content | Built in through X (Zapier) | Not their focus |
| Content policy | Terms allow "sexual situations" (Wired) | Refused prompts Grok accepted in Wired's test; Gemini made one of two images |
Sources
Changelog
Xiaomi MiMo is Xiaomi's AI model family and platform: chat in MiMo AI Studio, an OpenAI- and Anthropic-compatible API, voice models, and coding and desktop agents. Its MiMo-V2.6-Pro, released with MIT-licensed weights on September 22, 2026, is the top open-weights model on Artificial Analysis's index as of October 17, 2026.
Duck.ai is DuckDuckGo's private AI chat, launched in June 2024 as DuckDuckGo AI Chat. It lets you use models from OpenAI, Anthropic, Mistral and others without an account, and DuckDuckGo strips your IP address before each request. Six models are free; stronger ones come with DuckDuckGo's 9.99 or 19.99 USD monthly plans.
Gemini Notebook is Google's research notebook that answers from the sources you add and turns them into audio, video, reports and quizzes. Google renamed it from NotebookLM on July 16, 2026; it is the same standalone product, existing notebooks carried over, and notebooklm.google.com now redirects to notebook.google.com.
Character.AI is a chatbot app where people talk to characters other users create, made by Character Technologies. The company filed for Chapter 11 bankruptcy on October 12, 2026, and Disney bid 20 million USD for a license to some assets; the app runs during the sale and will be wound down if no bid keeps it going.
VideoEnhancer is a web tool that upscales short MP4 clips to 1080p, 2K or 4K with the FlashVSR model and also compresses, face-swaps and generates video. After a 5-second free trial, 1080p costs 3 credits a second, $0.03 to $0.09 depending on how credits are bought. VideoEnhancer is our own product: it is made by the team behind GenGrades.
LayerGrab splits a flat image into up to 16 named, transparent layers over a filled-in background, on the web, in Figma or in Chrome, or pulls out only what you name. Each layer costs $0.10 on its $12 plan, $0.08 on Premium. No independent review exists yet. LayerGrab is our own product: it is made by the team behind GenGrades.