Rent a GPU: cheapest hourly prices, and when buying wins
The cheapest on-demand price we found for each card across Vast.ai, Runpod and Lambda, next to RigFit's buy price, so you can see how many hours of use it takes before buying is cheaper.
Rental prices updated 2026-11-12 05:55 ET · buy prices Nov 12, 2026Cheapest rental per card
| GPU | Cheapest /hr | Break-evenhours until buying wins | |
|---|---|---|---|
| GeForce RTX 309024 GB | $0.14/hrVast.ai · marketplace | 11,315 hvs $1,618 to buy | RentVast.ai |
| GeForce RTX 409024 GB | $0.34/hrRunpod · Community Cloud | 9,018 hvs $3,066 to buy | RentRunpod |
| GeForce RTX 509032 GB | $0.50/hrVast.ai · marketplace | 12,569 hvs $6,247 to buy | RentVast.ai |
| RTX A600048 GB | $0.33/hrRunpod · Community Cloud | 13,545 hvs $4,470 to buy | RentRunpod |
| NVIDIA L40S48 GB | $0.60/hrVast.ai · marketplace | no RigFit buy price | RentVast.ai |
| NVIDIA A100 80GB80 GB | $0.87/hrVast.ai · marketplace | no RigFit buy price | RentVast.ai |
| NVIDIA A100 40GB40 GB | $0.40/hrVast.ai · marketplace | no RigFit buy price | RentVast.ai |
| NVIDIA H100 80GB80 GB | $1.99/hrRunpod · Community Cloud | no RigFit buy price | RentRunpod |
Break-even = RigFit's cheapest buy price ÷ cheapest hourly rental, ignoring electricity, resale value and the rest of the PC. Vast.ai = cheapest verified, on-demand 1-GPU offer with ≥95% host reliability (marketplace; includes default storage, excludes bandwidth). Runpod Community Cloud runs on vetted third-party hosts; Secure Cloud in Runpod's datacenter partners. Datacenter cards (L40S, A100, H100) have no tracked buy price. Rent links may be referral links.
Buy vs rent break-even calculator
Prefilled with today's cheapest buy and rental prices. Owning = (buy price − resale value) + electricity per hour of use. Renting = hourly price × hours. Electricity rate, power draw and resale value are yours to enter; we don't guess them.
Models that don't fit on one consumer GPU
Text models needing more than 32 GB (at Q4, else Q8) and the cheapest single rented GPU that holds them.
| Model | VRAM | Cheapest rentable card | From |
|---|---|---|---|
| Llama 3.3 70B Instruct Q4 | 43 GB | RTX A6000 | $0.33/hrRunpod |
| Qwen3 32B Q8 | 35 GB | RTX A6000 | $0.33/hrRunpod |
| Qwen2.5 Coder 32B Instruct Q8 | 35 GB | RTX A6000 | $0.33/hrRunpod |
| Mixtral 8x7B Instruct (MoE) Q8 | 48 GB | NVIDIA A100 80GB | $0.87/hrVast.ai |
| DeepSeek R1 Distill Qwen 32B Q8 | 35 GB | RTX A6000 | $0.33/hrRunpod |
More consumer cards by the hour
| GPU | Cheapest /hr | Break-even | |
|---|---|---|---|
| GeForce RTX 3060 12GB12 GB | $0.056/hrVast.ai | 6,071 h | Rent |
| GeForce RTX 3080 10GB10 GB | $0.11/hrVast.ai | 3,991 h | Rent |
| GeForce RTX 4060 Ti 16GB16 GB | $0.096/hrVast.ai | 6,875 h | Rent |
| GeForce RTX 407012 GB | $0.089/hrVast.ai | 9,348 h | Rent |
| GeForce RTX 4070 Super12 GB | $0.10/hrVast.ai | 6,796 h | Rent |
| GeForce RTX 4070 Ti Super16 GB | $0.15/hrVast.ai | 6,424 h | Rent |
| GeForce RTX 408016 GB | $0.20/hrVast.ai | 5,569 h | Rent |
| GeForce RTX 4080 Super16 GB | $0.20/hrVast.ai | 6,078 h | Rent |
| GeForce RTX 5060 Ti 16GB16 GB | $0.095/hrVast.ai | 7,579 h | Rent |
| GeForce RTX 507012 GB | $0.14/hrVast.ai | 5,882 h | Rent |
| GeForce RTX 5070 Ti16 GB | $0.16/hrVast.ai | 8,148 h | Rent |
| GeForce RTX 508016 GB | $0.19/hrVast.ai | 9,942 h | Rent |
FAQ
Is it cheaper to rent or buy a GPU for local AI?
It depends on hours of use. Divide the card's purchase price by the hourly rental price to get the break-even point in GPU-hours. For example an RTX 3090 at $1,618 to buy and $0.14/hour to rent breaks even after about 11,315 hours, before electricity and resale value.
Where do these rental prices come from?
Vast.ai's public marketplace search (cheapest verified on-demand single-GPU offer), Runpod's published pricing page (Community and Secure Cloud on-demand), and Lambda's published on-demand pricing. Refreshed daily; marketplace prices change by the minute.
What is the cheapest GPU to rent for a 70B model?
For Llama 3.3 70B at Q4 (~43 GB with an 8K context) the cheapest single card we found was a RTX A6000 at $0.33/hour on Runpod.
Sources (latest run)
- vastai: ok (23 prices)
- runpod: ok (18 prices)
- lambda: ok (5 prices)
Sources: Vast.ai public offer search, Runpod pricing, Lambda pricing. Prices are per GPU-hour in USD, before tax.
Get the weekly GPU price drops
Cheapest card per model, new and used. Free, 1 email a week.
Subscribe on beehiiv →