GPU cloud · head to head

Vast.ai vs Vultr Cloud GPU

What each costs, whether there's a free plan, and what each is best for, from prices checked on their own sites.

Vast.ai and Vultr Cloud GPU at a glance
Fact	Vast.ai	Vultr Cloud GPU
From	$0.49/hr	$1.67/hr
Free plan	No · No free credits	No · New-customer promo credit; GPU use not confirmed
Best for	A GPU rental marketplace where hosts set their own prices.	Cloud GPU instances in Vultr's global data centres.
Checked	Oct 17, 2026	Oct 17, 2026
Website	Visit	Visit

The difference in short

Vast.ai costs less to start: $0.49/hr against $1.67/hr for Vultr Cloud GPU.
Neither has a free plan; you pay from the start.

Vast.ai pricing

All about Vast.ai

Vast.ai plans and prices
Plan	Price	Specs / includes	Notes
RTX 4090 24GB	$0.49/hr	—	Marketplace, varies by host. Median $0.49 (feed: 0.4947); lowest offer 'from' $0.13 (0.1343); p10 0.4014; 437 offers at time of read. vCPU/RAM depend on the host.
RTX 5090 32GB	$0.64/hr	—	Marketplace, varies by host. Median $0.64 (feed: 0.6358); lowest offer 'from' $0.37 (0.3737); p10 0.4697; 572 offers.
L40S 48GB	$0.64/hr	—	Marketplace, varies by host. Median $0.64 (feed: 0.6406); lowest offer 'from' $0.43 (0.4274); p10 0.4281; 74 offers.
A100 80GB SXM4	$0.73/hr	A100 SXM4	Marketplace, varies by host. Median $0.73 (feed: 0.7347); lowest offer 'from' $0.43 (0.4271); 126 offers. A100 PCIE 80GB: median $0.73 (0.7337), lowest $0.37 (0.3676), 117 offers.
H100 80GB SXM	$2.40/hr	H100 SXM	Marketplace, varies by host. Median $2.40 (feed: 2.4007); lowest offer 'from' $1.73 (1.7337); 31 offers. H100 PCIE: median $2.47, lowest $2.34. H100 NVL: median $2.49, lowest $1.63.
B200 192GB	$9.38/hr	—	Marketplace, varies by host. Median $9.38 (feed: 9.3759); lowest offer 'from' $6.25 (6.2513); 33 offers.

Vultr Cloud GPU pricing

All about Vultr Cloud GPU

Vultr Cloud GPU plans and prices
Plan	Price	Specs / includes	Notes
L40S 48GB (1 GPU)	$1.67/hr	16 vCPU · 180 GB RAM	Same $1.671 per GPU for 1, 2, 4 and 8 GPU plans. 1-GPU plan: 16 vCPU, 180 GB RAM, 1.20 TB storage, 10 TB bandwidth.
A100 80GB PCIe (1 GPU)	$2.40/hr	A100 PCIe · 12 vCPU · 120 GB RAM	1-GPU plan: 12 vCPU, 120 GB RAM, 1.40 TB storage, 10 TB bandwidth. 8-GPU PCIe A100 node is $19.178/hr total. HGX A100 80GB SXM (8-GPU node) 'starting at' $2.800/GPU/hr.
A40 48GB (1 GPU)	$1.71/hr	24 vCPU · 120 GB RAM	1-GPU plan: 24 vCPU, 120 GB RAM, 1.40 TB storage, 15 TB bandwidth. 4-GPU plan $6.849/hr total.
HGX H100 80GB SXM (8-GPU node)	$2.30/hr	H100 SXM · 27 vCPU · 239.25 GB RAM	Per GPU, sold as an 8-GPU node (216 vCPU, 1914 GB RAM, 640 GB GPU RAM; vCPU/RAM above are the exact per-GPU share). The page's Cloud GPU section says 'pricing shown is on-demand', but the same $2.300 appears in the Bare Metal section where 'starting from' is stated to be for pre-paid contract terms. The public Vultr API (api.vultr.com/v2/plans) lists this plan at $23.92/hr per node ($2.99/GPU) with deploy_ondemand=false and a preemptible rate of $18.40/hr ($2.30/GPU). Treat $2.30 as a 'starting at' price and confirm the self-serve on-demand rate before quoting it as a flat price.
HGX B200 192GB (8-GPU node)	$3.50/hr	31 vCPU · 353.25 GB RAM	Per GPU, sold as an 8-GPU node (248 vCPU, 2826 GB RAM; per-GPU share shown). Same caveat as H100: the Bare Metal section says 'starting from' prices are for pre-paid contract terms. The public Vultr API lists the node at $68.00/hr ($8.50/GPU) with deploy_ondemand=false and a preemptible rate of $25.60/hr ($3.20/GPU), which does not match the page's $3.50, so the page figure and API disagree.

Other GPU cloud options

Cheapest first where the prices compare.

Compare all

FAQ

Which is cheaper, Vast.ai or Vultr Cloud GPU?

Vast.ai costs less to start: $0.49/hr against $1.67/hr for Vultr Cloud GPU. Prices checked Oct 17, 2026 on each company's own site.

Do Vast.ai and Vultr Cloud GPU have a free plan?

Vast.ai: No free credits. Vultr Cloud GPU: No free plan. New-customer promo credit; GPU use not confirmed.

What else should I compare with Vast.ai and Vultr Cloud GPU?

Other GPU cloud options we've checked: RunPod ($0.34/hr), TensorDock ($0.35/hr), DigitalOcean GPU Droplets ($0.76/hr), Modal ($1.10/hr).

Keep comparing

Built from the prices and facts we checked on each company's own site. We haven't tested either hands-on.How we work