Five companies now sell AI chatbot access on five different pricing ladders, and the gap between the cheapest paid tier and the most expensive one has widened to $292 a month. OpenAI, Anthropic, Google, xAI, and Perplexity each shipped new flagship models in September 2026, and each rewrote parts of its subscription lineup in the same window. That timing makes late September 2026 the first moment all five vendors’ current plans can be compared side by side using the same models: GPT-6 Sol, GPT-6 Luna, and GPT-6 Astra from OpenAI; Claude Opus 5.5 and Claude Fable 5.1 from Anthropic; Gemini 3 Pro from Google; and Grok 4.7 from xAI.
This breakdown pulls prices directly from each vendor’s own pricing pages and help documentation, cross-checked against dated 2026 reporting, so the numbers below reflect what a subscriber actually pays this month, not launch-day pricing from a year ago. Where a spec could not be confirmed on an official page, that gap is called out rather than guessed at.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
ChatGPT vs Claude vs Gemini vs Grok vs Perplexity: The Quick Comparison
Before diving into each vendor’s lineup, here is how the five services stack up on the features that actually change a subscriber’s day-to-day experience. Entry pricing spans from free to $8 a month, and top consumer tiers run from $99.99 to $300, a spread that makes “pick the $20 plan” bad advice for a lot of use cases.
| Feature | ChatGPT (OpenAI) | Claude (Anthropic) | Gemini (Google) | Grok (xAI) | Perplexity |
|---|---|---|---|---|---|
| Free tier | Yes, GPT-6 Luna | Yes | Yes | Yes (basic, via X) | Yes, 3 Pro Searches/day |
| Cheapest paid plan | Go: $8/month | Pro: $20/month | AI Pro: $19.99/month | X Premium: $8/month | Pro: $20/month |
| Mid-tier price | Plus: $20/month | Max 5x: $100/month | — | SuperGrok: $30/month | — |
| Top consumer tier | Pro: $200/month | Max 20x: $200/month | AI Ultra: $99.99/month | SuperGrok Heavy: $300/month | Max: $200/month |
| Newest flagship model (Sept 2026) | GPT-6 Astra (powers GPT-6 Pro) | Claude Opus 5.5 | Gemini 3 Pro | Grok 4.7 | Third-party model routing |
| Everyday consumer model | GPT-6 Sol / GPT-6 Luna | Claude Fable 5.1 | Gemini 3 Pro (Search AI Mode) | Grok 4 / Grok 4.7 (tier not itemized) | Multiple models, vendor-selected |
| Agent or coding tooling | Agent mode, Codex (40–400 msgs/month) | Claude Code included on Pro+ | Limited in consumer app | Not itemized publicly | Not a core feature |
| Image generation | Yes, all paid tiers | No dedicated image tool | Limited, tier-dependent | Grok Imagine, higher tiers | Yes, Pro and Max |
| Video generation | Not on current plan page | No | Veo access varies by tier | Grok Imagine, 1080p on Plus/Heavy | No |
| Deep research tool | Included on paid tiers | Not branded as such | Deep Search on AI Pro/Ultra | Not itemized publicly | Core Pro/Max feature |
| Bundled cloud storage | Not applicable | Not applicable | 20TB on AI Ultra | Not applicable | Not applicable |
| Annual billing discount | Business only ($20 vs $25/seat) | Pro drops to $17/month annual | Not published for US consumer tiers | Not published | Max: $2,000/year flat |
| Team/business seat price | $25/user/month ($20 annual) | $25–$125/seat/month | Handled via Workspace | Not a separate SKU | Not a separate SKU |
Free Tier Limits: What Each Vendor Gives Away Before You Pay
Every vendor in this comparison offers some form of free access, but the generosity of that free tier varies enough to change which paid plan actually makes sense as a next step. Perplexity publishes the hardest numbers here: 3 Pro Searches a day and exactly 1 Research query a month on Free, a cap tight enough that anyone doing weekly research work will hit it inside the first seven days. OpenAI and Google are vaguer, describing free access in terms of “standard” or “baseline” usage rather than a published daily or monthly count, which makes it harder for a free user to know in advance when they are about to get rate-limited.
| Vendor | Free Model Access | Published Usage Cap | Clear Upgrade Trigger |
|---|---|---|---|
| OpenAI | GPT-6 Luna (desktop app) | Not itemized as a fixed number | Wanting GPT-6 Sol or agent mode |
| Anthropic | Not itemized by model name | Baseline usage, described as lower than Pro’s 5x | Hitting session limits during heavy use |
| Standard Gemini access | Not itemized as a fixed number | Wanting Deep Search or extra storage | |
| xAI | Basic Grok via X | Not itemized as a fixed number | Wanting Grok Imagine or higher-tier Grok |
| Perplexity | Vendor-selected model | 3 Pro Searches/day, 1 Research query/month | Doing research more than once a month |
Perplexity’s transparency here cuts both ways. The hard numbers make it the easiest of the five to budget around, since a subscriber can calculate exactly how many Research queries they need and decide if $20 a month clears that bar. OpenAI, Anthropic, Google, and xAI all leave free-tier ceilings soft enough that the actual moment of hitting a wall, not a published spec, is what pushes most free users toward a paid plan.
ChatGPT Plus, Go, and Pro: What OpenAI Charges in 2026
OpenAI runs the widest ladder of the five vendors. Free users get GPT-6 Luna in the desktop app, a lightweight model OpenAI priced at $0.10 per million input tokens and $0.50 per million output tokens on the developer side, a hint at how cheap it is to serve. Go, OpenAI’s budget tier, lists at $8 a month in the United States, though rollout and pricing still vary by region since the plan started as a lower-cost pilot in Asian markets before expanding.
Plus remains $20 a month and is still the plan most subscribers land on. It unlocks GPT-6 Sol, the mid-weight model OpenAI shipped alongside Luna on September 22, 2026, at API pricing of $2 per million input tokens and $10 per million output tokens. Plus subscribers get voice mode, deep research, file and image tools, and agent mode, though OpenAI caps agent mode at 40 messages a month on this tier, a limit that pushes heavy agent users toward Pro.
Pro, at $200 a month, is the plan that actually unlocks GPT-6 Astra, the reasoning-focused model OpenAI debuted on April 9, 2026 and still labels its top-end offering as of late September. Pro subscribers get GPT-6 Pro, which OpenAI says is powered by Astra, plus Sol and Luna, and agent mode jumps to 400 messages a month. OpenAI’s help documentation describes ChatGPT Plus as providing “enhanced access to the ChatGPT web app for $20/month,” language that has held since Plus launched and still appears in OpenAI’s official support article.
Business pricing sits at $25 per user monthly, dropping to $20 per user with annual billing, matching Plus-level pricing at scale while adding admin controls and a default opt-out from model training on business data. Enterprise stays custom-quoted, as it has for every major AI vendor. One notable subtraction: OpenAI’s September 2026 pricing page no longer lists Sora video generation as a plan feature, following the wind-down of the standalone Sora app in April 2026 and the API’s retirement the same month Sol and Luna shipped.
Claude Pro, Max 5x, and Max 20x: Anthropic’s Price Ladder
Anthropic’s structure is more linear than OpenAI’s. Free costs $0. Pro costs $20 a month billed monthly, or $17 a month if you pay $200 upfront for the year, exactly the terms Claude’s official pricing page lists: “$17 per month with annual subscription discount ($200 billed up front). $20 if billed monthly.” That $3 monthly gap for committing a year in advance is bigger, proportionally, than the discounts OpenAI or Google publish for their consumer tiers.
Pro subscribers get roughly 5x the usage of Anthropic’s free tier per five-hour session and access to Claude Fable 5.1, the model Anthropic released on September 1, 2026. Anthropic’s original Claude Pro announcement framed the value proposition in near-identical terms years earlier, telling subscribers they would “gain 5x more usage of our latest model… for a monthly price of $20 (US),” a usage-multiplier structure the company has kept consistent as it renamed the underlying models.
Max comes in two sizes: Max 5x at $100 a month (5 times Pro’s usage per session) and Max 20x at $200 a month (20 times Pro’s usage). Both include Claude Code and Projects. Claude Opus 5.5, which Anthropic launched September 22, 2026 alongside OpenAI’s Sol and Luna, is priced at $4 per million input tokens and $20 per million output tokens on the API side, a figure that tells you how much heavier Opus is to run than Fable, even though both are reportedly available to Pro and Max subscribers.
Team plans split into Standard seats ($25/month, or $20 billed annually, at 1.25x Pro usage) and Premium seats ($125/month, or $100 annually, at 6.25x Pro usage). Enterprise starts around $20 per seat plus metered usage on top, according to Anthropic’s mid-2026 pricing documentation. Anthropic’s own support page keeps the description simple: “The Pro plan is available for $20 per month (US), with pricing in your local currency where supported.”
Google AI Pro vs Google AI Ultra: The Gemini Subscription Split
Google collapsed what used to be a confusing set of Gemini-branded plans into two paid tiers under the Google One umbrella: Google AI Pro at $19.99 a month and Google AI Ultra at $99.99 a month, both listed on Google One’s plans page. Google AI Pro gives subscribers higher-tier access to Gemini 3 Pro inside Search’s AI Mode, including Deep Search, Google’s answer to OpenAI and Perplexity’s research tools.
Google AI Ultra is the more interesting tier on paper. It bundles 20TB of cloud storage, a figure Google confirmed on its September 22, 2026 Gemini updates page, plus YouTube Premium at no added cost, and Google describes Ultra as delivering 5 times the usage limits of AI Pro. At $99.99 a month, Ultra undercuts both ChatGPT Pro and Claude Max 20x by exactly $100.01, while still bundling a consumer streaming perk neither rival offers.
Google is also the one vendor among the five that has not published a clear consumer-facing name for a next-tier model beyond Gemini 3 Pro. Gemini 3.8 Flash, which Google released on September 2, 2026 with introductory API pricing of $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026, is a developer-facing model rather than a confirmed Gemini app entitlement, so subscribers should not assume AI Pro or Ultra maps directly to it.
X Premium+, SuperGrok, and SuperGrok Heavy: xAI’s Tier Maze
xAI sells Grok access through two separate storefronts, and that split is the single most confusing thing about comparing these five services. Buy through X and you get X Premium ($8/month, basic Grok access bundled with the social platform) or X Premium+ ($40/month, which reportedly folds in SuperGrok access on top of X’s own premium features). Buy Grok standalone and the ladder resets: SuperGrok Lite at $10/month, SuperGrok at $30/month, SuperGrok Plus at $100/month, and SuperGrok Heavy at $300/month.
SuperGrok Heavy is the single most expensive consumer AI subscription among the five vendors covered here, and it reportedly throws in X Premium+ at no extra charge, a bundling move that only makes sense if you were already going to pay for both. SuperGrok Plus, at $100 a month, is notable for 1080p video generation through Grok Imagine, an entitlement 2026 reporting attaches specifically to that tier rather than the cheaper SuperGrok plan.
Grok 4.7 debuted September 21, 2026 as xAI’s current flagship, at API pricing of $2 per million input tokens and $6 per million output tokens. What xAI has not done, unlike OpenAI naming Sol for Plus and Astra for Pro, or Anthropic naming Fable for Pro and Max, is publicly map Grok 4.7 to a specific consumer tier. Subscribers paying $30 or $100 a month cannot currently confirm from xAI’s own materials whether they are running Grok 4 or Grok 4.7, a transparency gap none of the other four vendors share.
Perplexity Pro vs Max: Search-First Pricing
Perplexity keeps its lineup the simplest of the five: Free, Pro at $20 a month, and Max at $200 a month or $2,000 billed annually. Free accounts get 3 Pro Searches a day and one Research query a month, a hard cap that pushes anyone doing regular research work toward Pro within the first week of real use.
Pro adds Pro Search, Projects, file uploads, image generation, and a meaningfully higher usage ceiling, described by Perplexity’s own plan comparison as scaled to “average use” rather than a single fixed number. Max, at exactly the same $200 monthly price as ChatGPT Pro and Claude Max 20x, gives the highest limits Perplexity publishes and is aimed at subscribers running research or agentic workflows all day rather than occasional lookups.
What Perplexity does not clearly publish, and this matters if model transparency is a priority, is a locked-in list of which underlying models (GPT-6, Claude, Gemini, or Grok variants) each tier can route to. Perplexity’s help center maintains a models list for Pro and Max, but it functions more like a menu that changes than a fixed spec sheet the way OpenAI or Anthropic name their default models per plan.
Full Pricing Table: Every Plan, Monthly and Annual
Here is the complete list of every publicly listed consumer and team plan across all five vendors as of late September 2026, with monthly and annual pricing where a vendor publishes both.
| Vendor | Plan | Monthly Price | Annual Price | What It Unlocks |
|---|---|---|---|---|
| OpenAI | Free | $0 | — | GPT-6 Luna (desktop app) |
| OpenAI | Go | $8 | Not published | Higher limits than Free, GPT-6 Luna |
| OpenAI | Plus | $20 | Not published | GPT-6 Sol/Luna, agent mode (40 msgs/mo) |
| OpenAI | Pro | $200 | Not published | GPT-6 Pro (Astra), Sol, Luna, agent mode (400 msgs/mo) |
| OpenAI | Business | $25/user | $20/user (annual) | Sol/Luna/Astra, admin controls, training opt-out |
| OpenAI | Enterprise | Custom | Custom | Full model suite, enterprise security |
| Anthropic | Free | $0 | — | Baseline Claude access |
| Anthropic | Pro | $20 | $17/mo ($200 upfront) | Claude Fable 5.1, Claude Code, Projects, 5x Free usage |
| Anthropic | Max 5x | $100 | Not published | 5x Pro usage per session |
| Anthropic | Max 20x | $200 | Not published | 20x Pro usage per session |
| Anthropic | Team Standard | $25/seat | $20/seat (annual) | 1.25x Pro usage per seat |
| Anthropic | Team Premium | $125/seat | $100/seat (annual) | 6.25x Pro usage per seat |
| Anthropic | Enterprise | From $20/seat + metered | Custom | Metered usage, enterprise controls |
| Free | $0 | — | Standard Gemini access | |
| Google AI Pro | $19.99 | Not published (US) | Gemini 3 Pro in Search AI Mode, Deep Search | |
| Google AI Ultra | $99.99 | Not published (US) | 20TB storage, YouTube Premium, 5x AI Pro usage | |
| xAI | X Premium | $8 | Not published | Basic Grok access via X |
| xAI | SuperGrok Lite | $10 | Not published | Standalone entry-level Grok access |
| xAI | SuperGrok | $30 | Not published | Grok 4-tier access, higher limits than Lite |
| xAI | X Premium+ | $40 | Not published | SuperGrok access plus X Premium+ features |
| xAI | SuperGrok Plus | $100 | Not published | 1080p Grok Imagine video, higher usage, priority access |
| xAI | SuperGrok Heavy | $300 | Not published | Highest usage, multi-agent answers, X Premium+ included |
| Perplexity | Free | $0 | — | 3 Pro Searches/day, 1 Research query/month |
| Perplexity | Pro | $20 | Not published | Pro Search, Projects, file uploads, image generation |
| Perplexity | Max | $200 | $2,000/year | Highest limits, advanced research and agentic use |
Regional and Currency Pricing Differences
Every price quoted so far is the US dollar figure, and it is worth flagging that none of these five vendors charge a flat global rate. Anthropic’s own Claude Pro announcement listed the plan at “$20 (US) or £18 (UK),” a conversion that is not a straight currency swap since $20 converts to roughly £15-16 at typical 2026 exchange rates, meaning UK subscribers pay a real premium once local pricing and VAT are factored in rather than a pure currency conversion. OpenAI’s Go tier shows the widest regional spread of the five plans covered here: it started as a lower-cost pilot in Asian markets before expanding to the US at $8 a month, and OpenAI’s own materials note that availability and exact pricing still vary by region rather than following one global number.
Google and xAI both publish primarily US-dollar pricing on their main plan pages, with local-currency billing handled at checkout rather than listed upfront the way Anthropic explicitly calls out a UK price. Perplexity’s Max plan is the cleanest example of a vendor giving subscribers a predictable annual number regardless of billing cadence, at exactly $2,000 a year or $200 a month, a 1:12 ratio with no annual discount baked in, unlike Claude Pro’s $17-versus-$20 annual break or OpenAI Business’s $20-versus-$25 seat discount.
Business and Team Add-Ons: Where the Five Vendors Diverge Most
Individual pricing gets most of the attention, but the five vendors take genuinely different approaches once a team or company is paying the bill, and this is where the comparison matters most for IT buyers rather than individual subscribers. OpenAI and Anthropic both sell dedicated seat-based team products with published per-seat pricing: OpenAI Business at $25/user monthly ($20 annually) and Anthropic Team split into Standard ($25/seat monthly, $20 annually) and Premium ($125/seat monthly, $100 annually) tiers. That gives buyers an actual number to model against headcount before signing anything.
Google takes a different route, folding AI access into Google Workspace’s existing per-user licensing rather than selling a standalone AI seat, which means a company already on Workspace may find AI access cheaper to add than starting from scratch with a new vendor. xAI and Perplexity are the outliers here: neither currently sells a distinct team or business SKU separate from their consumer Pro/Max tiers, which makes them harder to deploy at company scale with per-seat billing, admin controls, and centralized invoicing the way OpenAI, Anthropic, and Google all support.
Data handling is the other divergence point that matters for business buyers. OpenAI’s Business tier explicitly excludes customer data from model training by default, a policy Anthropic and Google both apply similarly at their enterprise tiers, though the exact contractual language differs enough between vendors that IT and legal teams should read the current data processing addendum rather than assume parity based on marketing language alone.
Which Model Actually Powers Each Plan
OpenAI’s three-model September
OpenAI shipped three distinct GPT-6 variants inside six months: Astra in April, then Sol and Luna together on September 22-23. That is an unusually fast cadence even by 2026 standards, and it means anyone still comparing “ChatGPT” against Claude or Gemini without specifying which GPT-6 variant is talking is comparing the wrong thing. Sol is the general-purpose model Plus and Pro subscribers get for everyday work. Luna is deliberately cheap and lightweight, priced at a fraction of Sol’s API cost, and is the only model OpenAI hands to Free and Go users. Astra remains the reasoning-heavy model gated behind the $200 Pro tier and Business/Enterprise plans.
Anthropic’s Fable, Mythos, and Opus split
Anthropic released Claude Fable 5.1 on September 1, 2026 as the model reportedly available across Pro, Max, Team, and Enterprise. Claude Opus 5.5 followed three weeks later on September 22, priced far higher on the API side ($4/$20 per million tokens versus cheaper mid-tier models), and Anthropic’s public materials describe it as a more token-efficient successor to Opus 5, with reporting pegging the cost-per-task reduction at roughly 40% versus the prior Opus generation. Claude Mythos 5.1 shows up in Anthropic’s September 24 platform pricing documentation, but which consumer tier, if any, unlocks it is not clearly stated on the public consumer pricing page.
Google’s consumer app leans on Gemini 3 Pro for its higher-tier Search AI Mode experience, while Gemini 3.8 Flash exists mainly as an API product for developers rather than a named consumer-app entitlement. Grok 4.7 is confirmed as xAI’s current flagship as of September 21, 2026, but as noted above, xAI has not published which SuperGrok tier specifically runs it. Perplexity, by design, treats models as an interchangeable backend rather than a marketed feature, so its Pro and Max plans are sold on search and research capability rather than a named model.
Usage Limits and Benchmarks: What You Get for the Money
Raw price only tells half the story, since usage caps vary enormously between tiers even within a single vendor. OpenAI’s agent mode is the clearest published example: Plus subscribers get 40 agent messages a month, while Pro subscribers at ten times the price get 400, a 10x usage multiplier for a 10x price multiplier, at least on that one feature. Anthropic uses a similar multiplier logic across its entire lineup, framing every tier as a multiple of Pro’s baseline: Max 5x delivers 5 times Pro’s per-session usage, Max 20x delivers 20 times, and Team seats scale from 1.25x (Standard) to 6.25x (Premium).
Context window claims are the murkiest spec across all five vendors right now. A dated September 2026 third-party model comparison reports Claude running on a 200K-token context window across its paid consumer tiers, and separately reports OpenAI’s lower tiers splitting instant and reasoning modes at roughly 54K and 256K tokens respectively, with higher tiers reportedly reaching 128K instant and 400K reasoning. Neither figure is confirmed on Anthropic’s or OpenAI’s own consumer pricing pages as of this writing, so treat those numbers as third-party reporting rather than vendor-guaranteed specs, and check the current plan page before making a purchase decision based on context length alone.
Google’s usage framing is the most straightforward of the five: AI Ultra is simply 5 times AI Pro’s usage limits, full stop, with no per-feature breakdown published. That simplicity cuts both ways. It is easy to understand, but it also means subscribers cannot tell from the pricing page alone whether that 5x applies evenly across chat, image, and Deep Search, or whether one feature is disproportionately capped.
Five Real-World Scenarios: Matching a Plan to Your Workflow
Pricing pages describe features. They rarely describe which plan fits a specific job. These five scenarios reflect the workflows that most commonly determine which subscription actually pays for itself.
- A solo software contractor billing by the hour. Claude Pro at $20/month is the strongest fit, since Claude Code ships on the Pro tier and Fable 5.1 handles most day-to-day coding tasks without needing Opus 5.5’s higher API cost. Upgrading to Max 5x at $100/month only makes sense once agent-heavy workflows start hitting Pro’s session caps.
- A marketing team producing weekly video ads. SuperGrok Plus at $100/month is the only plan among the five that explicitly names 1080p video generation as an entitlement, via Grok Imagine, making it the most direct fit for a small team that needs short-form video output on a recurring schedule.
- A graduate student doing literature reviews. Perplexity Pro at $20/month solves the specific pain point Free cannot: one Research query a month is not enough for anyone doing regular academic work, and Pro Search plus file uploads directly targets citation-heavy research tasks.
- A household already paying for Google One storage. Google AI Ultra at $99.99/month effectively adds AI on top of a service many families already budget for, since the 20TB storage allotment and bundled YouTube Premium offset part of the AI-specific cost in a way none of the other four vendors replicate.
- A budget-conscious student or hobbyist. ChatGPT Go at $8/month or X Premium at $8/month are the cheapest paid entry points among the five vendors, and both beat Free-tier caps without jumping straight to a $20 plan.
- An operations team standardizing on one vendor. ChatGPT Business at $25/user monthly (or $20/user annually) or Claude Team Standard at $25/seat monthly are the two clearest team SKUs, since Google routes team access through Workspace and xAI and Perplexity do not sell a dedicated team seat product.
- A power user running agentic workflows all day. SuperGrok Heavy at $300/month is the highest-usage consumer tier of the five, reportedly bundling X Premium+ at no extra charge, making it the ceiling option for anyone who has already outgrown $200/month tiers elsewhere.
Pros and Cons of Each AI Subscription
ChatGPT. Pros: the widest tier ladder of the five, a genuine budget option at $8, and the clearest agent-mode message-count specs published anywhere in this comparison. Cons: the $20-to-$200 jump from Plus to Pro is the largest single price cliff among the five vendors, with no $100-ish middle tier for individual subscribers the way Anthropic offers Max 5x.
Claude. Pros: the most transparent usage-multiplier system (5x, 20x, 1.25x, 6.25x) of any vendor here, Claude Code bundled starting at Pro, and the biggest published annual discount for the entry paid tier ($17 vs $20). Cons: context window and computer-use specs are less clearly published on the official pricing page than third-party reporting suggests, and the free tier’s model access is not clearly itemized.
Gemini. Pros: only two paid tiers to choose between, a genuinely differentiated Ultra tier thanks to bundled storage and YouTube Premium, and the lowest sub-$100 flagship-adjacent price at $99.99. Cons: the least model-name transparency of any major vendor, since Gemini 3.8 Flash exists mainly as a developer product rather than a marketed consumer entitlement.
Grok. Pros: the cheapest true entry point via X Premium at $8, and the only vendor of the five to explicitly market 1080p video generation as a named plan feature. Cons: the most confusing dual-storefront structure of any vendor here, six named tiers across two purchase paths, and no public confirmation of which tier runs Grok 4.7 versus older Grok 4.
Perplexity. Pros: the simplest three-tier structure of the five, a Max annual price ($2,000/year) that at least gives heavy users a predictable yearly number, and a Free tier that, while limited, is still usable for occasional lookups. Cons: no named default model, a Free tier capped at one Research query a month that will frustrate anyone doing regular deep dives, and no team/business SKU distinct from Pro or Max.
Migrating Between AI Subscriptions Without Losing Your Work
Switching AI subscriptions, whether downgrading to save money or moving to a rival for a specific feature, follows roughly the same sequence across all five vendors. Skipping any of these steps is the most common reason people lose conversation history or paid credits during a switch.
- Export your conversation history first. ChatGPT, Claude, and Gemini all offer a data export option in account settings that produces a downloadable archive before you cancel anything.
- Check your billing cycle date, not just the calendar month. Most of these vendors bill on a rolling monthly date tied to signup, not the first of the month, so canceling a day early can still trigger a charge.
- Cancel through the same channel you subscribed with. A subscription bought through the Apple App Store or Google Play must be canceled there, not on the vendor’s website, or the charge will continue regardless of in-app cancellation.
- Re-create custom instructions, saved prompts, and Projects/Custom GPTs manually on the new platform, since none of the five vendors currently offer a direct one-click import from a competitor.
- Downgrade before you cancel if you are unsure. Moving from Pro to Plus, or from Max 20x to Pro, preserves your account and history while cutting cost, whereas a full cancellation on some plans resets usage history entirely.
- If moving for a specific feature (Claude Code, Grok Imagine video, Google’s storage bundle), confirm that feature is live on the destination tier today, not on a roadmap, since several of the entitlements in this comparison (Mythos 5.1’s tier mapping, Grok 4.7’s tier mapping) are not yet itemized by the vendor.
A simple way to track what you are exporting and re-creating is to keep a small checklist file alongside your downloaded export, so nothing gets lost between platforms:
migration_checklist = {
"conversation_export": "downloaded_2026-09-27.zip",
"custom_instructions": "copied to new_platform_notes.txt",
"saved_projects": ["Q4 report", "client-onboarding-prompts"],
"billing_cancel_date": "2026-10-15",
"cancel_channel": "vendor_website" # or "app_store" / "play_store"
}
What the Official Pricing Pages Say, Verbatim
Rather than paraphrase vendor pricing claims, it is worth quoting them directly, since the exact wording each company uses tells you how it wants a subscriber to think about value.
“ChatGPT Plus is a subscription plan that provides enhanced access to the ChatGPT web app for $20/month.”
OpenAI, official documentation (help.openai.com)
“The Pro plan is available for $20 per month (US), with pricing in your local currency where supported.”
Anthropic, official documentation (support.claude.com)
“With Claude Pro, subscribers can now gain 5x more usage of our latest model… for a monthly price of $20 (US) or £18 (UK).”
Anthropic, official announcement (anthropic.com)
“$17 per month with annual subscription discount ($200 billed up front). $20 if billed monthly.”
Anthropic, official pricing page (claude.com/pricing)
Notice that Anthropic’s own language, both in its original Claude Pro announcement and its current pricing page, leans on usage multipliers (“5x more usage”) rather than raw feature lists, a framing none of the other four vendors use as consistently across years of pricing pages.
The Verdict: Which AI Subscription Is Worth $20 (or More) in 2026
There is no single winner across five vendors selling fundamentally different products at different price points, but the data does point to clear defaults. For most individual subscribers who just want a reliable daily driver, Claude Pro and ChatGPT Plus remain the two safest $20/month picks, since both come with the clearest published usage terms and the newest consumer-facing models (Fable 5.1 and Sol, respectively) confirmed on their own pricing pages.
Google AI Ultra is the best value story of the five at $99.99/month, purely because it is the only tier bundling a large, independently useful benefit (20TB storage plus YouTube Premium) on top of AI access, meaning the effective AI-only cost is lower than the sticker price suggests for anyone who would buy that storage anyway.
xAI’s SuperGrok lineup is the hardest to recommend confidently, not because the pricing is bad, but because it is the only one of the five vendors that has not published which tier runs its newest flagship model, Grok 4.7. Until xAI closes that transparency gap the way OpenAI and Anthropic already have, SuperGrok is best suited to subscribers chasing a specific named feature, like Grok Imagine’s 1080p video, rather than subscribers chasing the newest underlying model.
Perplexity remains the specialist pick: skip it if you want a general chatbot, choose it if research and citation-heavy search is the actual job to be done, since its $20 Pro tier and $200 Max tier compete on search depth rather than raw model access.
Frequently Asked Questions
Is ChatGPT Plus or Claude Pro better value at $20 a month?
Both cost the same and both unlock a current-generation model (GPT-6 Sol and Claude Fable 5.1, respectively). Claude Pro includes Claude Code, making it the stronger pick for anyone doing regular development work, while ChatGPT Plus’s agent mode and broader tool integrations make it the stronger generalist pick.
Why is ChatGPT Pro $200 a month when Claude’s equivalent tier is also $200?
Both vendors settled on $200 as their top individual consumer price independently. ChatGPT Pro unlocks GPT-6 Pro (powered by GPT-6 Astra), while Claude’s $200 tier, Max 20x, is defined by a 20x usage multiplier over Pro rather than a distinct top-end model name.
What is the cheapest way to get a current-generation AI chatbot?
ChatGPT Go and X Premium are both $8 a month, the two lowest paid entry points among the five vendors compared here. Neither unlocks a top-tier model, but both beat the usage caps of a free account.
Does Google AI Ultra’s 20TB storage actually offset the $99.99 price?
If you already pay for a separate cloud storage plan of similar size, yes, since Google confirmed the 20TB figure alongside bundled YouTube Premium on its September 22, 2026 update. If you do not need extra storage, the effective AI-only value is harder to justify against Claude Max 5x at $100/month.
Which plan actually unlocks Grok 4.7?
xAI has not published a definitive tier mapping for Grok 4.7 as of late September 2026, despite the model launching September 21. Subscribers on SuperGrok or SuperGrok Plus cannot confirm from official materials alone whether they are on Grok 4 or Grok 4.7.
Is Perplexity worth paying for if I already have ChatGPT Plus or Claude Pro?
It depends on whether research and citation tracking is a regular task. Perplexity’s Free tier caps Research queries at one a month, so anyone doing that work weekly will hit the wall fast, but general chatbot use is already covered by an existing ChatGPT or Claude subscription.
Can I switch between these subscriptions without losing my conversation history?
Export your data first from account settings, since none of the five vendors offer a direct one-click import from a competitor. Downgrading within the same vendor (Pro to Plus, Max to Pro) preserves history; switching vendors entirely does not carry history over automatically.
Do any of these plans offer a student discount?
None of the five vendors publish a dedicated, named student discount tier as of September 2026 in the sources reviewed here. Business and Team seat pricing (OpenAI, Anthropic) is the closest published group-rate option, aimed at organizations rather than individual students.
Related Coverage
- ChatGPT vs Claude vs Gemini vs Grok: 20-Point IQ Gap [2026]
- ChatGPT vs Claude vs Gemini Memory: 2x Capacity Gap [2026]
- xAI Launches Grok 4.7 at $2/$6, Trails Claude by 7 Points [2026]
- Meta vs Google Stock: $2.3 Trillion Cap Gap [2026]
- OpenAI Admits AI Agents Leaked 53 ChatGPT Images [2026]
