Grok

What is Grok?

Grok is a general-purpose AI assistant developed by xAI; the official website also uses the SpaceXAI brand to showcase related products. It can be used on web pages, iOS, Android, and the X platform, and it offers capabilities such as dialogue, reasoning, programming, real-time web and X search, file analysis, voice processing, image and video generation, connectors, automation, and developer APIs.

Once a user logs in with the same account, conversations, settings, and SuperGrok subscriptions are synchronized between the Grok web page and the mobile app.

As of this verification, the main Grok model isGrok 4.5It should no longer be considered a chatbot that relies solely on X Dynamics.

The current product offers a comprehensive set of functions including Chat, Imagine, Voice, Build, Connectors, Automations, Business, and API.

Grok 4.5 model

Grok 4.5 is the current flagship model, designed primarily for intelligent programming, engineering, science, mathematics, and Agent workflows. Its API supports text and image input as well as text output; it features a context window of 500,000 tokens and enables reasoning, function calls, and structured output.

The model can think before providing an answer, and it can also access external information and systems through various tools.

The older Grok 4.3 and 4.20 versions are still available through the API: Grok 4.3 supports a context of 1 million tokens, with an emphasis on speed, reliability, tool invocation, and command compliance; version 4.20 offers reasoning, non-reasoning, and multi-Agent variants.

When making a choice, one should not consider only the version number; long documents, costs, latency, encoding capabilities, and tool invocation requirements all need to be taken into account.

Real-time web pages and X search

Grok can retrieve real-time content from public web pages and X, and use those search results to answer questions related to news, markets, individuals, and events. This is an important capability that sets it apart from purely offline models.

Users can continue to ask questions, request comparisons between multiple viewpoints, or verify the timeline.

Real-time does not mean absolute accuracy. X content may include rumors, satire, old images, advertisements, and misleading information, and web pages may also become outdated.

When dealing with news related to finance, healthcare, law, or public figures, it is necessary to consult the original sources, the date of publication, and the time when the event occurred; one should not rely on Grok’s summaries as the only source of evidence.

File and long document analysis

Grok supports uploading files such as PDFs, images, spreadsheets, code, and audio for summarization, extraction, Q&A, and analysis. Long-context models are suitable for reports, code repositories, and multi-file tasks, but very long inputs can increase usage and API costs.

Scanning PDFs, complex tables, low-resolution images, and data lacking context can still lead to parsing errors.

Connector

Connectors enable Grok to access users’ authorized email accounts, calendars, cloud storage services, collaboration tools, and business systems during conversations. The built-in connectors cover Gmail, Google Calendar, Google Drive, OneDrive, Outlook, Microsoft Teams, SharePoint, and Salesforce.

The catalog also offers services such as Box, Canva, Gamma, GitHub, Linear, Notion, and Vercel.

Developers can also connect to publicly accessible custom MCP servers.

When connecting to external accounts, the minimum required permissions should be used; OAuth authorizations should be checked regularly, and it is necessary to prevent the models from having unnecessary write permissions for high-risk operations. Enterprise administrators can configure connectors, roles, and organizational policies in a unified manner.

Grok Imagine images and videos

Grok Imagine supports the creation and editing of images from text or pictures, as well as multi-image editing. Its video-related functions include text-to-video, image-to-video, reference-based video generation, video editing, and various other capabilities; the specific interfaces vary depending on the version of the model.

Grok Imagine Video 1.5 improves the synchronization of motion, physics, sound, and lip movements, enabling it to generate images, sound effects, ambient noises, and dialogue all in one step; the Fast version takes about 25 seconds to produce a 6-second 720P video.

Imagine also supports project management, allows searching through the creation history, and enables the simultaneous activation of multiple agents to handle different creative tasks. AI-generated images and videos may contain errors related to identities, hands, text, physics, and sound; it is necessary to ensure that proper authorization exists before uploading real-person, brand, music, or film footage.

Voice mode and Voice API

Consumers can use hands-free voice conversations. The Voice API for developers offers capabilities for real-time voice-to-voice communication, speech recognition, speech synthesis, and the creation of custom voices.

Currently, the official version provides over 20 different multilingual voices, supporting more than 25 languages; expression can be controlled using pauses, whispers, and emphasis tags. Cloning a voice requires about one minute of audio, but explicit consent from the rights holder of that voice is necessary.

Automations – Automation

Grok Automation can be set to run once, daily, on weekdays, weekly, monthly, or annually; it can also be triggered when specific emails arrive. Users can use it to prepare summaries on a daily basis, monitor prices, or identify important items in emails, and they can choose to receive emails, notifications via app, or have only the results saved.

Each run creates a complete conversation that can be continued with further questions.

Scheduled automation is available to all users, while email-triggered automation is included in SuperGrok. Automation uses up usage quotas; when sending messages, making changes, or writing to external systems, manual confirmation and failure notifications are also required.

Grok Build programming agent

Grok Build is a terminal programming agent that can understand code repositories, edit files, run commands, search the web, and manage long-running tasks. It supports a full-screen terminal interface, headless scripts, and the Agent Client Protocol; it also allows for the expansion of skills, plugins, Hooks, MCP servers, and sub-agents. Installation packages are available for macOS, Linux, and Windows.

In July 2026, the developers will make the Agent framework and the terminal interface of Grok Build available under an open-source license. The code developed by the original team is licensed under Apache 2.0, while third-party code and the built-in code retain their respective licenses.

The framework can be compiled for use in local inference, but this does not mean that the weights of the Grok 4.5 model or the entire Grok service are made available under an open-source license.

Prices of the free version and SuperGrok

Starting in June 2026, the paid version of Grok will combine the usage amounts for Chat, Imagine, Voice, and Build into a weekly shared pool, with different tasks consuming varying percentages based on their computational load. Once the allocated amount is used up, users can wait for the weekly reset, upgrade their plan, or purchase additional usage.

The additional credit amount starts at $5, and it usually expires after one year from the time of purchase.

X Premium and SuperGrok are different subscription plans, with each platform being responsible for handling billing and refunds.

Grok API text pricing

The following shows the original price in US dollars per 1 million Tokens. Once the threshold of 200,000 Tokens for long context is reached, all Tokens in the same request will be charged at the rate applicable to long context.

• Grok 4.5：For short-context input, the cost is $2; for cached input, it is $0.30, and the output cost is $6. For long-context input, the cost is $4, the cached input cost is $0.60, and the output cost is $12.
• Grok 4.3：For short-context input, the cost is $1.25; for caching, it is $0.20; and for output, it is $2.50. For long-context input, the cost is $2.50, the caching cost is $0.40, and the output cost is $5.
• Grok Build 0.1：For short-context input, the cost is $1, the caching cost is $0.20, and the output cost is $2; for long-context input, the cost is $2, the caching cost is $0.40, and the output cost is $4.

Tools such as server-side web search, X-search, code execution, image generation, collection retrieval, and remote MCP are also charged based on the number of calls or additional resources used; therefore, the total cost of an Agent is not determined solely by the tokens.

Prices of Imagine and voice APIs

• Standard image:The cost for uploading an image is 0.002 dollars per image; for outputs in 1K or 2K resolution, the cost is 0.02 dollars per image.
• High-quality images:The cost for uploading an image is 0.01 dollar per image; the cost for a 1K resolution output is 0.05 dollars, and the cost for a 2K resolution output is 0.07 dollars.
• Imagine Image 2.0：Enter $0.01 per sheet; the output cost is approximately $0.04 to $0.08 per sheet, depending on clarity.
• Video 1.5：The cost for uploading an image is 0.01 dollar per image; the cost for outputting in 480P is 0.08 dollars per second, 0.14 dollars per second for 720P, and 0.25 dollars per second for 1080P.
• Real-time voice:For Think Fast 1.0, the cost is 0.05 dollars per minute; for version 2.0 it is 0.08 dollars per minute. Text input is charged separately.
• Speech recognition and synthesis:REST costs $0.10 per hour, streaming is $0.20 per hour, and text-to-speech synthesis costs $15 per million characters.

The API supports prepaid quotas and monthly billing; such quotas are usually non-refundable. In a production environment, it is necessary to set up team budgets, rate limits, key permissions, and cost monitoring.

Is it open source?

Grok Consumer Services and Grok 4.5 weights are not open-source products. The developers previously made the weights of Grok-1 available publicly, and in 2026 they made the Grok Build programming agent framework open source.

Neither of these two points can be used to prove that the current flagship model or the entire platform can be deployed in a private environment.

The catalog should indicate that “the main service is not open source, while Grok Build and some of the earlier projects are.”

Grok Usage Tutorial

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. Upload materials for which you have permission to use in Grok, or enter a search query;
3. First, a framework is established using the Grok 4.5 model, and then evidence is added through real-time web pages and X searches.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. A fixed set of research steps is established by combining the Grok 4.5 model, real-time web browsing, X search, as well as analysis of files and long documents.
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

Which users are it suitable for

• Users who need real-time information from websites and X for research, news, and market purposes.
• Professionals who need long documents, files, and data analysis.
• Programmers and engineering teams, especially for Agent coding and terminal workflows.
• Creators of image, video, and audio content.
• Individuals and businesses that wish to connect their email accounts, calendars, cloud storage services, and business systems.
• Developers who create search, Agent, media, and voice applications using APIs.

Restrictions and Precautions

• Grok may generate errors, biases, fabricated citations, or unsafe code, and real-time searches can also amplify unverified information;
• High-risk facts must be verified against their original sources;
• Before running the code and connectors, it is necessary to check the permissions, commands, and data ranges.
• Consumer usability, payment methods, and features also vary depending on the country, platform, and account.

Frequently Asked Questions

Is Grok free?

There is a free version that allows limited use of chatting, real-time searching, voice functions, and connectors. For higher limits and the premium features, SuperGrok is required.

What is Grok’s latest model?

The current main version is Grok 4.5, which has a context capacity of 500,000 tokens; it focuses on enhancing capabilities in coding, engineering, science, mathematics, and Agent tasks.

How much is SuperGrok?

SuperGrok costs $30 per month, while SuperGrok Plus costs $100 per month; regional taxes and the price of the mobile app may vary.

Can Grok generate videos?

Yes. Grok Imagine supports the creation of images and videos; Video 1.5 can generate videos with sound effects, background noises, and dialogue, and it offers an API resolution of up to 1080P.

Is Grok open source?

Currently, the Grok service and the 4.5 model are not open source. The Grok Build programming agent framework is licensed under the Apache 2.0 license; earlier versions of Grok-1 also had their weights made available publicly.