Cheap inference.

Mid-range GPUs that run quantized 7B–13B models at the lowest hourly rates.

GPU	Tier	Architecture	VRAM	Providers	AI models	From
Nvidia Nvidia GeForce RTX 3060	consumer	Ampere	12GB	2 →	—	$0.028/hr	Compare →
Nvidia Nvidia RTX A4000	workstation	Ampere	16GB	4 →	—	$0.042/hr	Compare →
Nvidia Nvidia RTX 5060 Ti	consumer	Blackwell	16GB	1 →	—	$0.056/hr	Compare →
Nvidia Nvidia GeForce RTX 4070	consumer	Ada Lovelace	12GB	2 →	—	$0.065/hr	Compare →
Nvidia Nvidia GeForce RTX 3090	consumer	Ampere	24GB	4 →	—	$0.071/hr	Compare →
Nvidia Nvidia RTX 5070	consumer	Blackwell	12GB	3 →	—	$0.082/hr	Compare →
Nvidia Nvidia GeForce RTX 4080	consumer	Ada Lovelace	16GB	3 →	—	$0.085/hr	Compare →
Nvidia Nvidia RTX 5070 Ti	consumer	Blackwell	16GB	2 →	—	$0.087/hr	Compare →
Nvidia Nvidia GeForce RTX 4090	consumer	Ada Lovelace	24GB	6 →	—	$0.11/hr	Compare →
Nvidia Nvidia GeForce RTX 5080	consumer	Blackwell	16GB	4 →	—	$0.11/hr	Compare →
Nvidia Nvidia RTX A5000	workstation	Ampere	24GB	4 →	—	$0.15/hr	Compare →
Nvidia Nvidia GeForce RTX 5090	consumer	Blackwell	32GB	5 →	—	$0.27/hr	Compare →
Nvidia Nvidia RTX A6000	workstation	Ampere	48GB	4 →	—	$0.29/hr	Compare →
Nvidia Nvidia RTX PRO 5000 Blackwell	workstation	Blackwell	48GB	2 →	—	$0.53/hr	Compare →
Nvidia Nvidia RTX 6000 Ada	workstation	Ada Lovelace	48GB	1 →	—	$0.56/hr	Compare →
Nvidia Nvidia RTX PRO 6000 WS	workstation	Blackwell	96GB	2 →	—	$0.66/hr	Compare →
Nvidia Nvidia RTX PRO 6000 S	workstation	Blackwell	96GB	1 →	—	$0.74/hr	Compare →
Nvidia RTX PRO 6000 Blackwell	workstation	Blackwell	96GB	1 →	—	$1.53/hr	Compare →