Grok AI: What it is, how it works and what’s new in 2026

Because model names and plan access change, I checked xAI’s model catalog and consumer pricing page on July 28, 2026. Grok AI combines a language model with optional web and X search, file analysis, voice, and media generation.

What is Grok AI?

Grok is xAI’s generative artificial intelligence assistant, available at grok.com, inside X, through the iOS and Android apps, and through the xAI application programming interface (API).

The assistant can answer questions, write and explain code, analyze uploaded files, create images and video, and hold voice conversations. Search access to public X posts, profiles, and discussions is useful when your task depends on a conversation unfolding on that platform.

How Grok AI works

Grok starts with a large language model (LLM). The model splits your prompt and conversation into tokens, processes the available context, and predicts a response token by token. Our guide to tokens in ChatGPT explains the same core mechanism shared by modern chat models.

The model does not know every event the moment it happens. xAI’s developer documentation says Grok 4.5 has a February 1, 2026 knowledge cutoff and needs Web Search or X Search for later information. Search results are added to the working context before the model writes its answer.

That distinction matters. A Grok answer based only on model knowledge can be stale, and a searched answer can inherit errors from webpages or X posts. Open the cited source before you rely on a claim about news, finance, health, law, or security.

Grok 4.5 is the flagship model in 2026

xAI’s model catalog lists Grok 4.5 as its flagship for chat, code, and agentic tool use. It has a 500,000-token context window, configurable reasoning, and separate search tools for information beyond its training cutoff.

Grok 4.5 reached grok.com, X, iOS, and Android on July 22, 2026, according to the consumer rollout announcement. The same model is available through the API under the grok-4.5 model name.

Older model names can remain in the API catalog for cost or workload reasons. Grok 4.3 has a larger one-million-token context window and lower short-context token prices, so the newest model is not automatically the cheapest choice for bulk document processing.

What Grok can do

The product groups several jobs in one interface. Availability and usage limits depend on your plan, account, and region.

• Chat and reasoning: explain a concept, compare options, draft text, or work through a multi-step problem.
• Web and X search: retrieve sources from the web and public conversations on X when the task needs recent information.
• File analysis: summarize and extract information from PDFs, spreadsheets, images, code, and audio files.
• Grok Imagine: generate or edit images and create video from supported prompts and inputs.
• Voice: talk to Grok through a spoken conversation in supported apps.
• Grok Build: use an agentic coding interface that can inspect files, edit a project, and run commands.

In July 2026, xAI also released workflows for Grok Build. A workflow divides a large coding or review task among focused agents, verifies their findings, and combines the result into one report. This is designed for work that splits cleanly, such as reviewing many routes for missing authorization checks.

A Google Workspace add-on released on July 24 puts Grok beside Sheets, Slides, and Docs. It can cite spreadsheet cells, create editable slides, and revise document content inside those apps.

Grok AI pricing

Consumer subscriptions and API billing are separate products. A SuperGrok subscription raises limits in the Grok apps, but it does not bundle API tokens for software you build.

Option	Published price	What it covers
Free	$0 per month	Limited consumer access to chat, search, voice, and connectors
SuperGrok	$30 per month	Grok 4.5, higher limits, Expert, connectors, image, and video generation
Business	$30 per user per month	Team billing, seat management, access controls, and business features
Enterprise	Custom	Custom infrastructure, identity, compliance, and support options
API	Usage based	Developer access billed by model tokens, tools, and media usage

xAI’s pricing comparison also lists SuperGrok Lite and SuperGrok Heavy, though the public table can vary in how it displays regional prices and limits. Check the official Grok pricing page from the account and region you plan to use before paying.

For developers, the official catalog lists Grok 4.5 at $2 per million input tokens and $6 per million output tokens for requests below the long-context threshold. Requests at or above 200,000 input tokens use higher rates, and search or code-execution tools add their own charges.

How to use Grok AI

The web and mobile apps are the simplest route for chat, research, files, voice, and media. Use X when your task starts with a post or public discussion, and use the API when Grok needs to run inside your software.

1. Open grok.com or install the official iOS or Android app.
2. Sign in and choose the tool your task needs, such as Search, Voice, or Imagine.
3. State the goal, relevant context, date range, and output format in your prompt.
4. For a searched answer, open the cited pages and separate source facts from social commentary.
5. Upgrade only after the free allowance interrupts a task you repeat.

For research about a fast-moving event, ask Grok to separate primary-source statements, reported facts, expert opinions, and unverified posts. That structure makes a live X discussion useful without treating popularity as evidence.

Grok AI vs ChatGPT

Both products support general chat, web research, files, coding, and media workflows. Grok deserves a closer look when public X discussions are central to the task, and ChatGPT may fit better when your existing workflow already depends on OpenAI’s apps or developer ecosystem.

Decision	Choose Grok when	Choose ChatGPT when
Research source	You need web sources plus public X posts	You need web research without an X-centered workflow
Coding workflow	You want Grok Build or xAI API models	You already use OpenAI coding tools or APIs
Media	You want Grok Imagine in the same product	Your team already uses OpenAI image or video tools
Procurement	xAI plan controls meet your requirements	OpenAI is already approved in your organization

The useful test is a task you already perform. Compare source quality, edit time, citations, and total cost on the same input instead of choosing from a benchmark or personality description. Our DeepSeek vs ChatGPT comparison applies the same workload-first decision to another pair of assistants.

Limits you should keep in view

Search access improves freshness but does not certify a source. X contains firsthand reports alongside jokes, speculation, edited media, and coordinated campaigns, so verify a cited post before treating it as evidence.

A large context window also does not guarantee that every detail receives equal attention. For long files, ask for citations to pages, sections, cells, or filenames, then inspect those locations yourself.

Do not paste passwords, API keys, private source code, customer records, medical data, or complete contracts into a consumer chat. If your team needs protected data, review the Business or Enterprise controls, retention terms, connector permissions, and contract before rollout.

Frequently asked questions

These answers use xAI’s consumer documentation, model catalog, and pricing pages checked on July 28, 2026.

A practical way to decide

Start with the free plan and run one task that depends on web or X sources. Keep Grok if its citations and source coverage reduce your research time after verification, and pay for a higher tier only when the limit blocks that repeatable workflow.