What AI subscriptions actually cost in India
The sticker price is rarely the price. A $20 plan bills closer to ₹2,320 once forex markup and 18% GST land on it — unless the provider has set a real rupee price, in which case it doesn’t. Work out the true cost of any plan, or of your whole stack.
These three settings change every rupee figure on this page. The defaults are stated assumptions, not a live feed — override them with what your own card statement says.
Every figure below is computed from the settings above. Filter by what you actually use AI for to bring the relevant plans to the top — the prices do not change, only the ordering.
| Model | Best at | Billing | Effective ₹/mo | Tier |
|---|
Almost nobody subscribes to just one. Tick everything you pay for — or are considering — and see the combined monthly and annual outflow in rupees, plus a flag when two of your picks overlap on the same job.
The ₹1,900 plan that bills you ₹27,846
A product designer in Pune signed up for a $20 AI subscription in January, did the arithmetic in her head at roughly ₹95 to the dollar, and budgeted ₹1,900 a month for it. The first card statement read ₹2,320. Nothing had gone wrong. Nobody had overcharged her. The plan simply costs what a dollar-billed digital service costs in India once the card issuer and the tax code have both taken their cut — and over a year that gap compounds to ₹5,046 she had not planned for.
This is the part almost every “best AI tool” listicle skips. They quote the dollar price, occasionally multiply by an exchange rate, and stop. But an Indian card paying a foreign software company passes through two levies that never appear on the provider’s pricing page: a foreign currency markup charged by your bank, typically between 2% and 3.5% depending on the card, and 18% GST on imported digital services, which the Indian tax code applies to exactly this kind of cross-border subscription.
Both are small percentages. Applied one on top of the other, monthly, they are not small money. And the second one — the 18% — is the one nobody can help you avoid.
The three prices of every AI subscription
Before comparing anything, it helps to separate three numbers that get carelessly used as if they were one.
The sticker price is what the provider’s website says: $20, $30, $10. It is the number in every comparison article and it is the least useful of the three for anyone paying from India.
The effective price is what actually leaves your bank account each month. For a dollar-billed plan that is the sticker price converted at your card’s rate, plus the issuer’s markup, plus 18% GST. For a plan with a genuine India SKU it is simply the rupee price on the page, because the provider has already set it locally with tax included. The calculator above computes this number, and it is the one the comparison table sorts on.
The cost per unit of work is what you actually get for the money, and it is the number almost nobody calculates. A ₹2,399 assistant you open forty times a week is cheap. A ₹1,740 video tool whose monthly credits buy four clips is expensive if you needed twelve. For chat assistants the seat is flat and this rarely bites. For creative tools it is the whole game, which is why the table shows each creative plan’s quota beside its price rather than pretending the monthly fee tells the story.
A worked example, line by line
Take a realistic freelance stack — a content designer running Claude Pro for writing and client documents, Canva Pro for design and cleanup, ElevenLabs for voiceover, and Runway for video clips. Here is what that costs from India, assuming a standard credit card at 3.5% forex markup and an exchange rate of ₹95 to the dollar. Both assumptions are adjustable in the calculator above; change them and every figure below moves with them.
| Plan | Listed price | How it is billed | Effective ₹/month |
|---|---|---|---|
| Claude Pro | ₹2,399 | India SKU, GST included, card only | ₹2,399 |
| Canva Pro | ₹499 | India SKU, UPI accepted | ₹499 |
| ElevenLabs Starter | $5 | USD → forex 3.5% → GST 18% | ₹580 |
| Runway Standard | $15 | USD → forex 3.5% → GST 18% | ₹1,740 |
| Total | — | — | ₹5,218/month · ₹62,622/year |
Most of the ₹5,218 carries no currency exposure at all. The sliver that does — ₹420 a month, or ₹5,046 a year — sits entirely on the two dollar-billed plans.
Avoidable: about ₹942 a year. A zero-forex card strips the 3.5% markup and the GST charged on top of it — real money, but modest.
Unavoidable: about ₹4,104 a year. That is 18% GST on the underlying dollar value, a tax on importing a digital service that no card or billing address changes.
Card optimisation is worth one month’s subscription a year. The bigger lever is whether the provider bothered to set an India price at all — when they do, the currency question disappears entirely.
Which AI plans actually have India pricing
Through 2025 and 2026 the major labs went from billing India in dollars to setting real rupee prices, and they did it at very different speeds and with very different generosity. This is now the single biggest determinant of what you pay.
| Provider | India price | UPI? | Notes |
|---|---|---|---|
| ChatGPT Go | ₹399/mo | Yes | Launched India-first in August 2025; roughly 10× the free tier’s limits. |
| ChatGPT Plus | ₹1,999/mo | Yes | GST included in the rupee price. |
| ChatGPT Pro | ₹19,900/mo | Yes | Power-user tier; rarely the right answer for individuals. |
| Google AI Plus | ₹399/mo | Yes | Introductory ₹199 for the first six months for new users. |
| Google AI Pro | ₹1,950/mo | Yes | Billed via Google Play; includes Veo video generation. |
| Claude Pro | ₹2,399/mo | No | Rupee pricing since 13 July 2026. ₹24,000/year works out to ₹2,000/month. Card and app-store billing only. |
| Claude Max | ₹11,999 (5×) · ₹23,999 (20×) | No | Same models as Pro, higher usage ceilings. |
| SuperGrok | ₹2,900/mo | No | Was ₹700 until July 2026, when xAI repriced India roughly to the global rate. |
| Canva Pro | ₹499/mo · ₹3,999/yr | Yes | The only mainstream creative tool with a real rupee price. |
| Perplexity Pro | USD only | — | No India SKU; forex and GST apply on the $20 price. |
₹399 has become India’s AI price point — OpenAI set it with ChatGPT Go, Google matched it with AI Plus. For most people one of those two is the right answer; the ₹2,000-plus tiers are capability they will not use.
The SuperGrok warning: India was the world’s cheapest SuperGrok market at ₹700 until July 2026, when xAI moved it to roughly ₹2,900 — about four times over, with the rupee barely moving in between. An India discount is a business decision, not a right, and it can be withdrawn in a single billing cycle.
The free layer nobody puts in comparison tables
All of which assumes you are paying. Before optimising a bill, it is worth checking whether you owe one.
Here is what a straight price comparison misses: for a large number of people the correct answer right now is ₹0, because their telecom operator is already paying.
Google partnered with Reliance Jio to give eligible users 18 months of Google AI Pro at no cost — a plan that otherwise runs about ₹1,950 a month. Eligibility: age 18+, active unlimited 5G plan, activated via MyJio. It is the single most valuable AI offer in the country, and much of the audience for ₹2,000-subscription articles is sitting on it unclaimed.
Airtel ran the same play first, with Perplexity: twelve months of Perplexity Pro, a subscription Perplexity itself valued at ₹17,000. That offer closed to new sign-ups on 16 January 2026. Activated before that date, you keep Pro until your term runs out — the earliest cohorts hit that wall around July 2026. Otherwise it is gone.
OpenAI ran the equivalent with ChatGPT Go, free for a year in India from August 2025. Reporting on the exact close date has been inconsistent — check your own OpenAI billing page rather than trusting any article, including this one.
The trap is not the free period — it is arriving at month thirteen with habits built around a tool you have never paid for, and discovering the renewal is ₹1,950 or ₹17,000 rather than zero. Calendar the expiry date the day you activate, and try the free tier or a ₹399 plan at least once during the free period so the eventual bill is a decision, not a default.
Creative tools price completely differently
Everything so far has assumed a subscription buys you a seat. For creative tools it usually doesn’t, and that changes the arithmetic enough to deserve its own section.
Chat assistants sell you a seat. Image, video, music and voice tools almost always sell you a credit or generation quota, and this changes how you should read their prices entirely.
Midjourney bills fast GPU time rather than a fixed image count, so its Basic tier runs out faster than people expect and most regular users end up on the higher tier with unlimited relaxed generations. Runway and Kling meter video by credits, where a month’s allowance on an entry plan buys only a handful of 1080p clips. Suno’s credits do not roll over from month to month. ElevenLabs meters by characters of audio generated, so long-form narration burns through an allowance far faster than short social clips do.
The practical consequence: before subscribing to any creative tool, work out your realistic monthly volume first and check it against the quota second. A ₹1,740 plan that covers your output is cheaper than a ₹950 plan you exhaust in week two and top up twice. The comparison table above shows each creative plan’s quota next to its rupee price for exactly this reason.
One India-specific note: essentially every creative tool except Canva Pro bills in dollars. Canva’s ₹499 rupee plan with UPI covers background removal, photo cleanup, everyday design and passable slide decks. It is not best-in-class at any single one of those jobs, which is a real trade-off — but it is by some distance the cheapest way to cover most of them from India without touching a forex-billed card.
When capability outweighs the price gap
Everything so far has argued that the rupee price is the number to watch. Here is the honest limit of that argument.
The flagship assistants cluster within a few hundred rupees of one another once you have applied the India-SKU rule. At that spread, price stops being a useful tie-breaker — the difference between the cheapest and dearest frontier plan is roughly one coffee a month, and capability differences between them are real rather than marketing.
Claude leads on complex multi-file coding and long documents, which is why it sits underneath Cursor and Windsurf. Gemini leads on graduate-level science and maths and handles images and video natively. Perplexity is the only mainstream assistant showing an inline citation for every claim — for research, journalism and legal work that is the entire product, not a feature. Grok is alone in having live access to what is being posted on X. ChatGPT is the broadest generalist and the safest single choice.
Which means the sequence matters. Fit decides which plan; the pricing rules in this page decide what that plan costs you. Choosing Perplexity because you need citations and then paying the forex-and-GST premium is a defensible decision. Choosing Perplexity because a comparison table put it near the top, and never noticing it has no India SKU, is not.
And below a certain usage level neither question applies. Under roughly ten to fifteen queries a day, the free tiers of ChatGPT, Claude, Gemini and DeepSeek will all serve you — and the elaborate task-matching above becomes a way of choosing which excellent thing to pay for unnecessarily. Establish that you need to pay before optimising what you pay.
Six ways to pay less — and what each one costs you
Every one of these works. Every one also has a downside cost-cutting articles tend to skip.
1. Prefer plans with an official India SKU
The largest single lever, worth 15–20% before you even look at the tier. Trade-off: it narrows your choice to providers who bothered. If Perplexity’s citations are what you need, no amount of preference makes it cheaper — you pay the premium or use a worse tool.
2. Pay by UPI where it is offered
UPI sidesteps the card forex markup entirely; OpenAI, Google and Canva all support it. Trade-off: Anthropic does not. Claude has rupee pricing but still needs a card, so the best coding assistant here is the one you cannot pay for the easy way.
3. Use a zero-forex card for dollar-billed plans
Removes the 2–3.5% markup and the GST charged on top of it. Trade-off: per the worked example, that is roughly ₹942 a year on a ₹2,320-a-month load — genuine, but one month’s subscription, not a transformation.
4. Take annual billing where the discount is published
Claude Pro falls from ₹2,399 to an effective ₹2,000 a month annually; Canva and Gamma publish annual rates too. Trade-off: it locks in a price for a year, in a market where a leading model can be replaced in a few months.
5. Actually use the free tiers
For light use, the free tiers of ChatGPT, Claude, Gemini and DeepSeek are genuinely sufficient. Trade-off: rate limits arrive mid-task, and the China-based free models carry a data-residency question that matters for client-confidential work.
6. Stop stacking overlapping tools
One plan matched to your main task usually beats two that overlap — the stack builder above flags the duplication. Trade-off: your one assistant will be second-best at something you occasionally do; an aggregator like Poe is the middle path.
Edge cases most comparisons miss
The web price and the app store price are not always the same. Providers list separately on claude.ai, the iOS App Store and Google Play. Check both before a team purchase, where the gap multiplies by seat count.
If you are GST-registered, the 18% may not be a real cost. A registered business sits in a different tax position from an individual consumer. Case-specific — a question for your CA, not a calculator.
Free tiers on China-based models carry a data-residency question, not a quality one. DeepSeek, Kimi, Qwen and MiniMax are strong models; the question is where the data goes. Irrelevant for a textbook summary, not for a client’s financials. The table flags this rather than ranking it down, because the right answer depends on what you upload.
Credits do not always roll over. On several creative tools an unused allowance expires at month end. Lumpy work — three heavy weeks, one quiet one — costs more per output than the plan page implies.
A promotional price is not a price. Google AI Plus at ₹199 becomes ₹399 after six months. A telco bundle ends. SuperGrok went up fourfold. Budget on the renewal figure, not the acquisition one.
What to actually do
The designer in Pune from the opening was not careless. She did the only arithmetic the internet had given her — multiply the dollar price by the exchange rate — and it was wrong by ₹5,046 a year. The whole point of this page is that the correct arithmetic is not harder, just less widely published.
So take the sequence rather than any single number:
- Check whether you are already covered. Jio users should check MyJio for the 18-month Google AI Pro bundle before paying for anything at all.
- Establish that you need to pay. Run a fortnight honestly on a free tier. If you never hit a limit, stop here — the rest of this is optimisation of a cost you do not have.
- Let capability pick the plan. Within the frontier tier the price spread is small and the capability gaps are real, so choose on fit and accept the resulting bill.
- Then let pricing pick how you buy it. Prefer a rupee price and UPI where the fit is close, because that is where the 15–20% sits — not in your choice of card.
- Add up the whole stack, not the one plan. The number that matters is the annual figure in the burn-rate card, and it is usually larger than people expect.
- Diary every renewal and promo expiry the day you subscribe, because a promotional price is not a price.
Two of those six steps are about the tool you choose. Four are about what happens to the money afterwards — which is roughly the ratio the rest of the internet has backwards.
The sources this tool is built on
Every price, rate and threshold above traces to a provider’s own pricing page or to contemporaneous reporting of a specific announcement. Where sources conflict or a promotional status is unclear, it is flagged as such rather than smoothed into a confident number.
Methodology: Plans with an official India SKU are shown at the provider’s published rupee price, which already includes GST; no forex or additional tax is applied to them. Plans billed in USD are converted at the exchange rate you set (default ₹95), then loaded with the foreign-currency markup you select (default 3.5%), then with 18% GST on the marked-up amount — the order matters, because GST applies to the converted transaction value. Annual pricing is applied only where the provider publishes a specific annual rate; plans without one hold at the monthly figure rather than displaying an assumed discount. Free tiers are shown at ₹0 with their material limitations noted rather than being ranked as if unconstrained. The stack builder sums effective monthly costs under your current settings and flags overlaps where two selected plans serve the same primary task.
Scope and limits: This tool models list pricing for individual consumer plans. It does not model team or enterprise seats, API token spend, GST input credit for registered businesses, telco or student bundle eligibility, or promotional pricing whose availability depends on your specific account. Task recommendations reflect published capability differences and are editorial judgement, not a scored ranking — there is no composite score behind them, because any such score would be invented. AI subscription pricing changes faster than almost any other consumer category; treat every figure here as accurate to its verification date and confirm on the provider’s own page before you subscribe.
