AI Compare
Prices for United States, before sales tax Data release 2026.10.16-3, Oct 16, 2026 Rules 2.1.0
Comparing Claude Pro and X Subscription Premium: individual plans, prices for United States. Change the plans below to compare any other two.
The verdict is being reviewed because facts about these plans changed. The comparison below already shows the current facts.
This page is for developers deciding between Claude Pro from Anthropic ($20 a month, or $200 a year paid up front) and Cursor Premium ($30 a month). Claude is a general assistant that includes Claude Code, Anthropic's coding agent. Cursor is a code editor with agents, sold since October 12, 2026 as the X Subscription, which also includes Grok and X Premium. We compare these two plans because each is the lowest of its product that documents cloud agents that open pull requests and automatic pull request review. We also cover Claude Free and Max, and Cursor Free and Plus, where they change the choice.
Everything here comes from Anthropic's, Cursor's and xAI's own pages, checked on October 12 and 13, 2026. Where a company does not say something, we say it is not documented, which is not the same as saying it is missing. We have not tested either tool yet, so nothing below is a claim about the quality of their code or their reviews.
$20 a month or $200 a year
$30 a month or $300 a year
Opus, Sonnet, Haiku; Fable via usage credits.
Latest Grok models (Grok 4.7) and Composer 2.5 from the shared pool; third-party models (Claude Opus 5.5, Claude Fable 5.1, GPT-5.6 Sol, Gemini 3.1 Pro, etc.) available but billed separately at API list price via top-ups.
Up to 1M on all individual plans.
Grok 4.7: 256k standard window, 500k long context (input over 256k billed at 2x). Pay-per-use third-party models go higher, e.g. Claude Opus 5.5 up to 1M tokens.
Pricing comparison lists only Anthropic models for Pro: Fable (uses 50% of weekly limits), Opus, Sonnet, Haiku. No models from other AI companies are offered.
Pick models from SpaceXAI (Grok), Cursor (Composer), Anthropic, OpenAI, Google and Meta (Muse Spark); third-party models are not in the included pool and are billed at API list price.
Claude chat on web, desktop and mobile for writing, editing, questions and everyday work (at least 5x Free usage per 5-hour session); chat and Claude Code share one usage pool.
Cursor itself is coding-only, but this X Subscription tier bundles Grok Chat, a general-purpose chatbot, sharing the same usage pool.
Web search on all plans.
Agent has a built-in Web tool that runs web searches (Agent is ticked on every plan). Whether answers show source links is not documented.
Research.
No multi-step research-report mode documented for Cursor plans; Projects coordinate many agents on code work, not research reports.
PDF, DOCX, CSV, TXT, XLSX etc.; 500MB per file, up to 20 files per chat (no plan split stated).
Agent can read image files (.png, .jpg, .gif, .webp, .svg) into chat for vision-capable models; iOS app lets you attach photos or files. PDF/document upload into chat is not documented.
Create and edit files with code execution.
Agent can run terminal commands (so it could run scripts), but spreadsheet/CSV analysis is not documented as a feature.
Claude Code included; 5-hour session + weekly limits.
Cursor Agent (edits files, runs terminal commands, searches code) ticked on every plan; Free is limited, paid tiers scale usage (Plus base, Premium 4x, Super 15x, Ultra 50x Plus).
Claude Code extension for VS Code and Cursor (also other VS Code forks such as Devin Desktop, Kiro) and plugin for JetBrains IDEs (IntelliJ IDEA, PyCharm, Android Studio, WebStorm, PhpStorm, GoLand). Shares Pro usage limits.
Own editor (Cursor); Cursor agent inside IntelliJ IDEA, PyCharm, WebStorm and other JetBrains IDEs via ACP (2025.1+, uses your Cursor plan); Xcode 26.3+ projects via MCP bridge.
Claude Code is an agent (chat panel, inline diffs) in VS Code/JetBrains; no page documents inline as-you-type code completions. The Desktop app docs explicitly say it has no autocomplete; the IDE extension docs are silent.
Tab autocomplete (multi-line, cross-file, next-edit jumps). xAI says Premium includes Cursor Pro benefits, and Cursor Pro listed unlimited tab completions; the new pricing page itself states no Tab limit.
Claude Code on the web: cloud sessions run on Anthropic-managed VMs, keep running after you close your laptop, clone GitHub repos and create PRs; auto-fix PRs. Counts toward Pro usage limits; no separate VM charge.
Cloud Agents & automations: run in cloud VMs on GitHub/GitLab/Bitbucket/Azure DevOps repos, work on a branch and open PRs; start from web, Slack, GitHub comments, iOS. Uses shared pool.
Via Claude Code GitHub Actions authenticated with a Pro subscription token: review workflow posts inline PR comments automatically. Managed Code Review (research preview) is Team/Enterprise only; local /code-review also works.
Bugbot reviews GitHub/GitLab pull requests automatically on each update, comments with bugs, security issues and fix suggestions. Included from Premium; usage-based.
No photo/illustration generation; can build diagrams/charts/SVG visuals (beta, web/desktop).
Grok Imagine image generation bundled in The X Subscription; Cursor Agent also has a tool that generates images into the project's assets/ folder. Draws on the shared usage pool.
Claude Motion (beta) marked No on Free, Pro, Max.
Grok video generation is bundled from Premium up (pricing table tick); draws on the shared usage pool.
Voice mode on all plans.
Via bundled Grok Voice (not inside the Cursor editor; the Cursor iOS app offers dictation only). Uses the shared pool.
Anthropic says Claude does not generate photos or illustrations; it can view and analyze uploaded images. No page documents editing an uploaded photo.
Grok Imagine (bundled with The X Subscription) edits images: Image 2.0's magic wand edits the region you point at, plus segmentation, multi-image editing (up to 5 inputs) and a Photo Edit template. Draws on the plan's shared Cursor and Grok usage pool.
No Anthropic page documents background removal. Claude does not generate photos or illustrations.
Grok Imagine Image 2.0 has background removal that exports a subject with a transparent background, and a 'BG Removal & Change' template. Draws on the plan's shared Cursor and Grok usage pool.
No image generation: Claude does not make photos or illustrations, so there are no generated images to put text in. It builds HTML and SVG diagrams, charts and visuals instead.
xAI says Imagine Image 2.0 plans typography and layout so small text comes out sharp (Quality Mode on grok.com/imagine and the apps). Draws on the plan's shared Cursor and Grok usage pool.
No logo maker or logo template is documented. Claude Design (all plans) makes visuals, mockups, prototypes, one-pagers and landing pages.
No logo template is documented for Grok Imagine; its templates include Mascot Maker, Icon Maker and Merch Maker, none described as a logo maker.
Custom visuals (beta, web and desktop, all users) are built with HTML and SVG; you can download a visual as an .svg or .html file.
No xAI or Cursor page documents SVG or vector export from Grok Imagine.
No Anthropic page documents image upscaling for Claude.
No xAI page documents image upscaling for Grok Imagine; 'Smart resize' changes the aspect ratio by filling in the frame, which is not described as upscaling.
Claude Motion (short animations built as code from your text, charts and images, exported as MP4) is marked No on Free, Pro and Max; it is a Team and Enterprise beta and uses no video generation model.
Bundled Grok Imagine (grok.com/imagine, iOS, Android) animates an image into video, up to 1080p, with up to 7 image references; draws on the shared weekly usage pool.
No Claude page documents editing uploaded video footage on Free, Pro or Max; Claude Motion edits only its own code-built animations and is not on these plans.
Grok Imagine video editing is documented only for the xAI API (billed separately, not from the plan pool); no official page documents footage editing in the Grok apps or Cursor.
No Claude page documents a digital presenter. Claude Motion (not on this plan) 'doesn't create realistic footage or people'.
No Cursor or xAI page documents a digital presenter in the bundled Grok apps.
No Claude page documents lip sync on any plan.
Grok Imagine generates dialogue with synced speech, but no official page documents lip syncing existing video to new audio.
No Claude page documents automatic video captions on any plan.
No Cursor or xAI page documents automatic video captions.
No Claude page documents turning long videos into short clips on any plan.
No Cursor or xAI page documents turning long videos into short clips.
Claude makes no video on this plan (Claude Motion is marked No on Free, Pro and Max), so a video watermark does not apply.
Grok Imagine images and videos carry a visible Grok watermark; there is no setting to remove it.
Artifacts (all plans) turn a description into single-page websites, small interactive tools and apps that can call Claude; Claude Design makes prototypes and landing pages. Claude Code (included in paid plans) builds full apps in your codebase.
Grok Build in Grok (web, iOS, Android) builds an app, game, website or dashboard from a description, live in the chat. Grok Build ticked on this plan; uses the shared usage pool.
Artifacts start private; Pro and Max can share with 'Anyone with the link' (viewers sign in to Claude; not for artifacts that use Claude or connectors). Claude Code artifacts publish to a claude.ai URL that can be public with no sign-in; one self-contained page, no backend.
Apps you publish from Grok Build get their own grok.me address; you choose just you, anyone with the link, or the whole internet. Grok Build ticked on this plan; uses the shared usage pool.
No Anthropic page documents serving an artifact on your own domain. Legacy published artifacts offer embed code with an Allowed domains list, which is embedding, not a custom domain.
Grok Build lets you point a domain you own at a published app.
Artifacts can store data between sessions: personal or shared storage, 20 MB per artifact, text only. Visitors of AI-powered artifacts sign in with their own Claude account. No other database or auth service is documented.
No xAI page documents a database or user sign-in for Grok Build apps; it describes apps 'with real state and logic', secrets for third-party API keys, and connectors for dashboards.
Claude Design exports a .zip, standalone HTML, or hands off to Claude Code or tools such as Lovable, Replit, v0 and Vercel; Claude Code (included in paid plans) writes code in your own project, and published artifact pages come from a local file.
Grok Build can push the whole project to a GitHub repo to continue in an editor or in the Grok Build terminal agent; Cursor itself edits code in your local project.
No official Claude page describes generating songs or any audio output.
No official Cursor or Grok page found that describes generating songs with vocals; the plan table covers Grok Chat, Imagine (images, video from Premium) and Voice.
No official Claude page describes generating instrumental music.
No official Cursor or Grok page found that describes generating instrumental music.
No official Claude page describes stem downloads.
No official Cursor or Grok page describes stem downloads.
Voice mode lets Claude talk back in a spoken conversation (pricing table: Voice mode Yes), but no official page describes turning text you supply into a voiceover file.
Text to Speech exists in the Grok Voice API suite ($15 per 1M characters), not as a plan feature. Paid plans include Grok API access, but API usage is billed separately and does not draw from the plan pool.
No official Claude page describes copying a voice.
Voice Cloning exists in the Grok Voice API suite (clone from two minutes of speech, 30 free clones), not as a plan feature. Paid plans include Grok API access, but API usage is billed separately and does not draw from the plan pool.
No official Claude page describes translating and re-voicing audio or video.
Grok Voice lists 'Live Translation' (speak in one language, be heard in another) as Coming Soon; no plan includes dubbing today.
No official Claude page describes generating sound effects.
No official Grok or Cursor page describes generating stand-alone sound effects from a description.
The upload help article lists document types (PDF, DOCX, CSV, TXT and more) and images only; no audio or video file formats are listed and transcription is not described.
Grok chats accept audio (MP3, WAV, M4A, OGG, FLAC, AAC) and video (MP4, MOV) uploads up to 150 MB and can transcribe and interpret them; quality varies. Uses the shared weekly Grok and Cursor usage pool.
No official page describes Claude joining or recording meetings. Connectors (for example Microsoft 365) can read existing meeting notes or summaries.
No official Cursor or Grok page describes joining or recording meetings and writing notes.
Dictation is available to all plans, but only inside the Claude iOS and Android apps to speak prompts; no official page describes typing into other apps.
No official page describes system-wide dictation; the Cursor iOS app offers dictation inside the app only (per earlier catalog research).
Projects.
Cursor Projects (beta): a coordinator agent per repo runs cloud agents, keeps synced shared-context files across months; plus version-controlled project rules in .cursor/rules on every plan.
No GPTs/Gems-style assistants listed; Skills and Projects listed separately.
Custom subagents (markdown files in .cursor/agents with own prompt, tools and model), plus skills and plugins; usable in editor, CLI and Cloud Agents. No plan restriction stated.
Memory across conversations.
Project-scoped only: Projects keep synced files where agents record what they learn about the codebase and your preferences; Bugbot can save '@cursor remember' facts as rules. No general cross-chat memory documented.
Directory connectors (e.g. Google Drive, Gmail, Google Calendar, GitHub, Slack, Microsoft 365) + custom remote MCP. Also Claude for Microsoft 365.
MCP servers (e.g. Google Drive, Notion) connect Agent to external tools; installed from Customize or mcp.json. Slack/Linear/GitHub etc. integrations mainly drive Cloud Agents (Premium+).
No Claude add-on that works inside Gmail is documented. Claude in Chrome (a browser agent) can act on Gmail pages, and the Gmail connector works from Claude's own chat; neither is an integration inside Gmail.
No official page describes a Cursor or Grok add-on inside Gmail. Cursor's Gmail plugin (Aug 3, 2026) and Grok's Gmail connector work from Cursor or Grok chat, not inside Gmail. Grok for Google Workspace lists Docs, Sheets and Slides only.
Claude for Google Workspace add-on: Claude sidebar in Google Docs (Extensions > Claude). Beta; Chrome, Edge, or Safari (not Firefox); Google account must be allowed to install Marketplace apps.
Grok for Google Workspace (Jul 24, 2026) works inside Google Docs as a free add-on, but no official page says which plans can use it; the Workspace Marketplace listing only says 'Free of charge'. Cursor itself has only a connector-style plugin.
Claude for Google Workspace add-on: Claude sidebar in Google Sheets (Extensions > Claude). Beta; Chrome, Edge, or Safari (not Firefox); Google account must be allowed to install Marketplace apps.
Grok for Google Workspace (Jul 24, 2026) works inside Google Sheets as a free add-on, but no official page says which plans can use it; the Workspace Marketplace listing only says 'Free of charge'. Cursor itself has only a connector-style plugin.
Claude for Word add-in, generally available. Word on the web, Windows (Microsoft 365, Version 2205+), and Mac (16.61+). Pricing table: Claude for Microsoft 365 Yes.
Grok for Word (bundled Grok) is a free add-in, but its Microsoft Marketplace listing names only legacy plans: SuperGrok, Heavy, Business and Enterprise. No official page maps that to The X Subscription Plus, Premium, Super or Ultra.
Claude for Excel add-in, generally available. Excel on the web, Windows (Microsoft 365 subscription), and Mac (16.46+); not Excel on iPad. Pricing table: Claude for Microsoft 365 Yes.
Grok for Excel (bundled Grok) is a free add-in, but its Microsoft Marketplace listing names only legacy plans: SuperGrok, Heavy, Business and Enterprise. No official page maps that to The X Subscription Plus, Premium, Super or Ultra.
Claude for PowerPoint add-in, generally available. PowerPoint on the web, Windows (Microsoft 365 subscription), and Mac (16.46+). Pricing table: Claude for Microsoft 365 Yes.
Grok for PowerPoint (bundled Grok) is a free add-in, but its Microsoft Marketplace listing names only legacy plans: SuperGrok, Heavy, Business and Enterprise. No official page maps that to The X Subscription Plus, Premium, Super or Ultra.
Claude for Outlook add-in, Beta. Needs Exchange Online through Microsoft 365 and a one-time tenant-wide Microsoft Graph consent by a Global Administrator; not Outlook on iOS or Android. Pricing table: Yes.
Grok for Outlook (bundled Grok, Jul 21, 2026) runs beside your inbox: summarizes threads, drafts replies, triages mail. For all paid X and SuperGrok users; X Subscription paid plans include X Premium benefits once you connect your X account.
Claude in Chrome; hand off and schedule tasks.
Agent's Browser tool controls a browser to navigate, click and screenshot (for testing your app). No general computer-use agent documented.
User chooses 'Model Improvement' at signup/in-app; changeable anytime in Privacy Settings. Default not stated.
Privacy Mode in settings: code data not used for training by Cursor or model providers (ZDR agreements). With Privacy Mode off, Cursor may use codebase data and prompts to train.
Export from web or desktop (not iOS/Android).
No chat-history or account data export documented on the pages checked.
Ad-free commitment.
No ads documented inside Cursor. The bundled X component shows timeline ads by tier: Free none removed, Plus 'Less timeline ads', Premium+ ad-free on 'most surfaces'.
Conflict: Anthropic's help center says you own outputs and may use them in your products; the Consumer Terms we could read say non-commercial use only. Check the terms for your account.
Cursor terms assign Suggestions (generated code) to you; SpaceXSI consumer terms (Grok) say you keep ownership of User Content, which includes Output, but using Output with its name or brand needs permission and attribution.
No cloud-storage bundle.
No general cloud storage. Premium+ includes 'Code hosting via Origin' (Cursor's git forge, early beta); no storage quota stated.
Web, iOS, Android, desktop.
Cursor desktop editor, Cursor CLI (macOS/Linux/WSL/Windows), web agents at cursor.com/agents, iOS/iPadOS app (docs list paid plans), JetBrains IDEs via ACP; Grok apps and X via the bundle.
Claude Pro is Anthropic's general assistant, with Research, web search, voice, projects, memory, file uploads and connectors such as Google Drive, Gmail, GitHub and Slack. Claude Code shares one usage pool with chat. Claude does not generate photos or illustrations on any plan.
Cursor describes itself as a coding agent. Premium adds cloud agents and automations, Bugbot, Origin code hosting (early beta) and Grok Bot, and xAI says it bundles SuperGrok, Cursor Pro and X Premium+. The general assistant in the bundle is Grok Chat, with Grok Imagine for images, Grok Voice and, from Premium, Grok video generation. All of it draws on the pool you also use for code.
On Pro, Claude Code runs as an extension for VS Code and Cursor (and other VS Code forks such as Devin Desktop and Kiro), and as a plugin for JetBrains IDEs including IntelliJ IDEA, PyCharm, Android Studio, WebStorm, PhpStorm and GoLand. Completions as you type are not verified: the extension docs are silent, and the Desktop app docs say that app has no autocomplete.
Cursor is its own desktop editor, with a CLI for macOS, Linux, WSL and Windows, web agents at cursor.com/agents and an iOS and iPadOS app listed for paid plans. Its agent also runs in JetBrains IDEs 2025.1 or later through ACP, on your Cursor plan, and in Xcode 26.3 or later projects through an MCP bridge. Premium documents Tab autocomplete with multi-line, cross-file and next-edit jumps. The new pricing page states no Tab limit; xAI says Premium includes Cursor Pro benefits, which listed unlimited Tab.
Claude Code on the web runs cloud sessions on Anthropic-managed virtual machines that keep running after you close your laptop, clone GitHub repos, create pull requests and offer auto-fix for pull requests, within Pro usage and with no separate machine charge. Repos hosted elsewhere are not documented on the pages we checked. Cursor's cloud agents run in cloud machines on GitHub, GitLab, Bitbucket or Azure DevOps repos, work on a branch and open pull requests, started from the web, Slack, GitHub comments or the iOS app.
For reviews, Claude Pro uses Claude Code GitHub Actions authenticated with your Pro subscription token: a review workflow posts inline comments on pull requests automatically. Anthropic's managed Code Review (research preview) is for Team and Enterprise only. Cursor's Bugbot reviews GitHub and GitLab pull requests automatically on each update and comments with bugs, security issues and suggested fixes; it is usage-based. To see what each review catches, open a pull request with the bug from our code repair kit, and give both agents the same task from the code spec kit.
Claude Pro includes Anthropic models only: Opus, Sonnet and Haiku, with Fable through usage credits (on Max, Fable can use up to 50% of the weekly limits). Context goes up to 1 million tokens. Limits are a session limit that resets every five hours plus a weekly limit. Max 5x ($100 a month) and Max 20x ($200) give five and twenty times Pro usage per session, are monthly only, and include $100 or $200 a month in Claude Platform API credits.
Cursor Premium gives 4 times Plus usage in a pool that refreshes weekly, shared by Cursor and Grok. Grok 4.7 and Composer 2.5 come from that pool; Grok 4.7 has a 256K standard window and 500K long context, with input over 256K billed at 2x. Models from Anthropic, OpenAI, Google and Meta, such as Claude Opus 5.5, GPT-5.6 Sol and Gemini 3.1 Pro, are outside the pool and billed at API list price through top-ups, with spending caps (see API prices).
Anthropic lists the Claude Code extension for Cursor on Pro and Max, and it shares your Claude plan's usage limits; the pages we checked do not say which Cursor plan, if any, it needs. Inside Cursor itself, Claude Opus 5.5 and Claude Fable 5.1 are among the third-party models from Premium, billed at API list price in addition to the $30 subscription. Cursor Free has no third-party models and Plus has Grok models only. Whether Max's API credits can pay for Claude models in Cursor is not documented.
Claude Pro is $20 a month, or $200 a year on the web paid up front (about $17 a month); whether tax is included is not documented, and Anthropic points mobile buyers to their app store's price. Cursor Premium is $30 a month before tax, monthly only.
Anthropic asks you to choose the Model Improvement setting at sign-up and lets you change it in Privacy Settings, but does not state the default. Cursor's Privacy Mode keeps your code out of training by Cursor and model providers, under zero data retention agreements; with it off, Cursor may train on codebase data and prompts. Claude data can be exported from the web or desktop app; no export is documented for Cursor. Anthropic states an ad-free commitment on every plan. No ads are documented inside Cursor; the X part of Premium is ad-free on "most surfaces".
Claude Free includes Sonnet and Haiku, web search, file work with code execution, voice and up to five projects, with rolling five-hour limits, but no Claude Code, editor integration, cloud sessions or Research. Cursor Free gives limited agent and chat requests, a starter allowance for Cursor's coding agents and limited Grok, with no cloud agents, Bugbot or third-party models; its Tab limit is not stated. Plus, an $8 introductory plan, adds Basic usage with Grok models only.
Cursor Premium. It documents Tab autocomplete with multi-line, cross-file and next-edit jumps. For Claude Code, no page we checked documents completions as you type, and the Claude Desktop app docs say it has no autocomplete, so this is not verified for Claude.
Cursor Premium. Its cloud agents document all three plus GitHub, and Bugbot reviews GitLab pull requests too. Anthropic documents Claude Code cloud sessions that clone GitHub repos; other hosts are not documented on the pages we checked.
Claude Pro. Opus, Sonnet and Haiku come within its usage limits, and the Claude Code extension also runs inside the Cursor editor. On Cursor Premium, Claude models cost extra at API list price.
Claude Pro includes Research, file analysis, connectors and voice, without ads. Cursor Premium bundles Grok Chat, Grok Imagine images and Grok video, sharing one pool with your coding. Claude does not generate photos or illustrations.
Claude Pro: $200 a year on the web, paid up front, about $17 a month. Cursor's pricing page shows monthly prices only, so Premium is $30 every month.
Both document automatic reviews. Bugbot on Cursor Premium reviews GitHub and GitLab pull requests on each update, billed by usage. Claude Pro runs a GitHub Actions review workflow with your subscription token. Test both with our code repair kit.
We have not run these tests for you unless a result above says so. Each kit gives you the same files, the same prompt and the same checks for both tools.
From Premium ($30 a month) up, yes, but they are not in the included pool: Cursor bills them at API list price through top-ups, with spending caps. Cursor Free has no third-party models, and Plus has Grok models only.
Anthropic documents a Claude Code extension for VS Code and Cursor on Pro and Max, using your Claude plan's limits. Claude Free does not include Claude Code. Which Cursor plan, if any, this needs is not documented on the pages we checked.
Since October 12, 2026, Cursor's plans are sold as the X Subscription, and Cursor is now owned by SpaceX. xAI says Premium bundles SuperGrok, Cursor Pro and X Premium+, and Cursor and Grok draw on one pool that refreshes weekly.
Not verified. Claude Code is documented as an agent with a chat panel and inline diffs. The extension docs do not mention completions as you type, and the Desktop app docs say it has no autocomplete. Cursor Premium documents Tab completions.
Claude asks you to choose the Model Improvement setting and does not state the default. Cursor's Privacy Mode keeps code out of training by Cursor and model providers; with it off, Cursor may train on codebase data and prompts. Check both.
Neither company's documentation answers that, and we have not tested either yet. Run our code repair, code spec and code refactor kits in Claude Code and in Cursor with the same files and the same prompt.
Every fact links to its source and shows when it was verified. Read how AI Compare works and the change log.