01000001 01010000 01001001QWEN-3.8-MAXDEDICATEDQWEN-3.7-PLUS01001110 01000110 01011001QWEN-3.8-27BFRONTIERDEEPSEEK-V4-FLASHGLM-5.2NO-FOREX-MARKUP01000001 01010000 01001001QWEN-3.8-MAXDEDICATEDQWEN-3.7-PLUS01001110 01000110 01011001QWEN-3.8-27BFRONTIERDEEPSEEK-V4-FLASHGLM-5.2NO-FOREX-MARKUP
DOC. APF-009·GPU RENTALS·LIVE

GPUs by the hour.
Killed by your wallet.

// vs a USD-billing gateway, all-in
19.6%
Same weights. Same throughput.

Buying the same model in dollars, you pay a 5–7% forex markup and 18% GST on the conversion before the model has run. We hold the provider accounts and bill you rupee-native, so neither exists on your bill.

₹95.5 interbank × 1.06 forex × 1.18 GST vs ₹96.0 flat
▸ See the full comparison

One rupee-priced account across 30+ GPU clouds — we find the cheapest card that fits and rent it for you. The part that matters: your prepaid balance is a hard ceiling. When it runs out the instance is destroyed, not billed. You cannot wake up to a bill you never agreed to.

§ 01 — GUARANTEES
// The four that matter before the price does

You cannot be surprised by this bill.

Renting a GPU is easy. Not being billed for one you forgot about is the hard part, and it is the part that is designed here rather than promised.

[01] CEILING
Your wallet is the ceiling
The meter runs against a prepaid balance and nothing else. There is no card on file, no overdraft and no invoice at the end of the month. When the money is gone the machine is destroyed — the worst case is a stopped instance, never a bill you did not agree to.
[02] EXPIRY
Everything expires
Every rental is started with a hard lifetime, not an open-ended session. It ends when that lifetime ends whether or not anyone remembers it exists. A forgotten instance is a bounded loss by construction.
[03] IDLE
Idle means stopped
An instance nobody is using is an instance nobody should be paying for. Idle rentals are stopped automatically rather than left running politely in the background.
[04] TEARDOWN
Teardown is verified
We ask the provider whether the machine is really gone before we tell you it is, and the answer goes on a signed receipt. If the provider has not confirmed it, the page says so plainly instead of showing you a green tick — an unconfirmed teardown may still be billing.
§ 02 — WHERE IT RUNS
// You choose, before you start, not after

India, or the cheapest card anywhere.

The picker below carries an India-only filter, and every offer in the grid says which it is. Pick before you rent: the region is on the quote and on the receipt.

◱ INDIA REGION
Keep it in India
India-resident region available
Filter the catalogue to India offers and rent one. It is priced separately from international capacity and it is usually not the cheapest card available — the trade is yours to make, with both numbers in front of you.
◲ INTERNATIONAL
Cheapest card that fits
Region named on your quote
Where the cheapest cards actually are. We make no India-residency claim for it and we do not bury it either.
Most capacity is international spot; the region is on your receipt before you start, not after. Spot capacity is cheap because it can be reclaimed — that is the trade, and it is why every rental has a hard lifetime and a verified teardown rather than an assumption that the machine will still be there tomorrow.
§ 03 — PICK A GPU
// Live catalogue · 30+ clouds · rupee quote before you commit

The cheapest card that fits your job.

Rates come from the live catalogue at the moment you ask, so there is no rate card here to go stale. Pick a family, pick an offer, see the quote — nothing starts until you confirm it.

Rent a GPU

WALLET-LINKED AUTO-KILL

Billed by the minute from your wallet while it runs. It stops when the money runs out, when its lifetime expires, or the second you press TERMINATE — and we ask the provider whether the machine is really gone before we tell you it is.

GPUs
picked: — none yet —
dies automatically after 6h, or when the wallet hits zero

Rent it. Forget it. Safely.

Prepaid rupees, a hard lifetime, and a signed receipt for every teardown.

▸ Get your API keyDeploy a model instead