FastGPU
Live-GPU-Cloud-Preisvergleich über 28+ Anbieter (RunPod, Vast.ai, Lambda, Hyperscaler). Gehosteter MCP-Server über Streamable HTTP, keine Installation und kein Schlüssel erforderlich: Ein Agent fügt einfach die URL https://fastgpu.co/api/mcp hinzu. Tools: list_gpu_prices und match_workload.
Dokumentation
FastGPU: find the cheapest live GPU cloud for your AI workload
FastGPUCompareProvidersAPIPricingBlogLaunch a GPU
Live
VAH100 SXM$1.34/hr▲49%TMI300X$1.71/hr▲39.8%CCH100 SXM$1.79/hr▲49%VPH100 SXM$1.99/hr▲49%RH100 PCIe$1.99/hrLGH200$2.29/hrPMI300X$2.59/hr▲39.8%TCH100 PCIe$3.20/hrCMI300X$3.45/hr▲39.8%AH100 NVL$6.98/hr▲7.8%TV100$0.19/hrHL40$1.00/hrVL40S$1.67/hrCRTX 4090$0.39/hrVAH100 PCIe$1.87/hrTMI325X$2.25/hrCCL40S$0.87/hrTAH200$2.99/hr▼24%GCH200$2.60/hr▼24%DH200$2.69/hr▼24%PMI325X$3.80/hrTCA6000$0.35/hrGCB200$11.28/hrBB200$9.98/hrTA6000$0.45/hrHA100 80GB$1.35/hr▲50.1%VA40$1.71/hrTMI355X$2.85/hrCCA100 80GB$1.50/hr▲50.1%TAB200$4.09/hrRH100 NVL$2.59/hr▲7.8%JLA30$0.41/hr
VAH100 SXM$1.34/hr▲49%TMI300X$1.71/hr▲39.8%CCH100 SXM$1.79/hr▲49%VPH100 SXM$1.99/hr▲49%RH100 PCIe$1.99/hrLGH200$2.29/hrPMI300X$2.59/hr▲39.8%TCH100 PCIe$3.20/hrCMI300X$3.45/hr▲39.8%AH100 NVL$6.98/hr▲7.8%TV100$0.19/hrHL40$1.00/hrVL40S$1.67/hrCRTX 4090$0.39/hrVAH100 PCIe$1.87/hrTMI325X$2.25/hrCCL40S$0.87/hrTAH200$2.99/hr▼24%GCH200$2.60/hr▼24%DH200$2.69/hr▼24%PMI325X$3.80/hrTCA6000$0.35/hrGCB200$11.28/hrBB200$9.98/hrTA6000$0.45/hrHA100 80GB$1.35/hr▲50.1%VA40$1.71/hrTMI355X$2.85/hrCCA100 80GB$1.50/hr▲50.1%TAB200$4.09/hrRH100 NVL$2.59/hr▲7.8%JLA30$0.41/hr
updated just now·28 providers·40 GPUs
·cheapest A100 40GB$0.27/hr
Compare live cloud GPU prices across the entire market.
Describe your workload and get the live market ranked for that job: the cheapest GPU that actually fits, in stock right now, across marketplaces, neoclouds and hyperscalers.
Find GPUs
Serve DeepSeek V3.2 to 1000 usersFine-tune Qwen3 32BCheapest H100 in the EUGenerate video with Wan 2.28x H100 with InfiniBand for trainingRun Whisper transcription cheaply
The live GPU market, matched to your workload. Real prices, real availability, no guessing.
Product
Guides
- GPU Cloud Price Index
- Best GPU by workload
- Rent GPUs by the minute
- Rent GPUs by the second
- Free GPU price dataset
Legal
Contact
© 2026 FastGPU. All rights reserved.
Live GPU prices
Cheapest live price per GPU
By GPUAll offersBy provider
GPU
VRAM
Cheapest / hr
Cheapest at
Offers
[
B300
DDeepInfra
NVIDIA · Blackwell Ultra
288 GB
$4.89/hr
On-demand
DDeepInfra
15
](https://fastgpu.co/gpus/b300)[
MI350X
PPaperspace
AMD · CDNA4
288 GB
$6.16/hr
On-demand
PPaperspace
2
](https://fastgpu.co/gpus/mi350x)[
GB300
VVerda
NVIDIA · Blackwell Ultra
288 GB
$8.62/hr
On-demand
VVerda
2
](https://fastgpu.co/gpus/gb300)[
MI355X
TTensorWave
AMD · CDNA4
288 GB
$2.85/hr
On-demand
TTensorWave
2
](https://fastgpu.co/gpus/mi355x)[
MI325X
TTensorWave
AMD · CDNA3
256 GB
$2.25/hr
On-demand
TTensorWave
4
](https://fastgpu.co/gpus/mi325x)[
B200
DDeepInfra
NVIDIA · Blackwell
192 GB
$3.69/hr
On-demand
DDeepInfra
23
](https://fastgpu.co/gpus/b200)[
MI300X
TTensorWave
AMD · CDNA3
192 GB
$1.71/hr
On-demand
TTensorWave
5
](https://fastgpu.co/gpus/mi300x)[
GB200
GCGMI Cloud
NVIDIA · Blackwell
192 GB
$8.00/hr
On-demand
GCGMI Cloud
3
](https://fastgpu.co/gpus/gb200)[
H200
VAVast.ai
NVIDIA · Hopper
141 GB
$1.98/hr
On-demandIn stock
VAVast.ai
26
](https://fastgpu.co/gpus/h200)[
GH200
LLambda
NVIDIA · Grace Hopper
96 GB
$2.29/hr
On-demand
LLambda
2
](https://fastgpu.co/gpus/gh200)[
H100 NVL
VAVast.ai
NVIDIA · Hopper
94 GB
$2.27/hr
On-demandIn stock
VAVast.ai
5
](https://fastgpu.co/gpus/h100-nvl)[
H100 SXM
VAVast.ai
NVIDIA · Hopper
80 GB
$1.34/hr
On-demandIn stock
VAVast.ai
31
](https://fastgpu.co/gpus/h100-sxm)[
H100 PCIe
VAVast.ai
NVIDIA · Hopper
80 GB
$1.87/hr
On-demandIn stock
VAVast.ai
5
](https://fastgpu.co/gpus/h100-pcie)[
RTX PRO 6000
VAVast.ai
NVIDIA · Blackwell
96 GB
$0.94/hr
On-demandIn stock
VAVast.ai
19
](https://fastgpu.co/gpus/rtx-pro-6000)[
A100 80GB
VAVast.ai
NVIDIA · Ampere
80 GB
$0.55/hr
On-demandIn stock
VAVast.ai
27
](https://fastgpu.co/gpus/a100-80gb)[
L40
VAVast.ai
NVIDIA · Ada Lovelace
48 GB
$0.34/hr
On-demandLow stock
VAVast.ai
9
](https://fastgpu.co/gpus/l40)[
RTX 6000 Ada
RRunPod
NVIDIA · Ada Lovelace
48 GB
$0.74/hr
Community
RRunPod
5
](https://fastgpu.co/gpus/rtx-6000-ada)[
L40S
RRunPod
NVIDIA · Ada Lovelace
48 GB
$0.79/hr
Community
RRunPod
17
](https://fastgpu.co/gpus/l40s)[
A6000
RRunPod
NVIDIA · Ampere
48 GB
$0.33/hr
Community
RRunPod
10
](https://fastgpu.co/gpus/a6000)[
A40
RRunPod
NVIDIA · Ampere
48 GB
$0.49/hr
On-demand
RRunPod
2
](https://fastgpu.co/gpus/a40)[
RTX PRO 5000
RRunPod
NVIDIA · Blackwell
48 GB
$0.82/hr
Community
RRunPod
2
](https://fastgpu.co/gpus/rtx-pro-5000)[
A100 40GB
VAVast.ai
NVIDIA · Ampere
40 GB
$0.27/hr
On-demandIn stock
VAVast.ai
14
](https://fastgpu.co/gpus/a100-40gb)[
RTX 5000 Ada
RRunPod
NVIDIA · Ada Lovelace
32 GB
$0.49/hr
Community
RRunPod
1
](https://fastgpu.co/gpus/rtx-5000-ada)[
RTX PRO 4500
RRunPod
NVIDIA · Blackwell
32 GB
$0.72/hr
On-demand
RRunPod
3
](https://fastgpu.co/gpus/rtx-pro-4500)[
L4
VAVast.ai
NVIDIA · Ada Lovelace
24 GB
$0.08/hr
On-demandLow stock
VAVast.ai
9
](https://fastgpu.co/gpus/l4)[
A5000
VAVast.ai
NVIDIA · Ampere
24 GB
$0.15/hr
On-demandIn stock
VAVast.ai
3
](https://fastgpu.co/gpus/a5000)[
A10
AAWS
NVIDIA · Ampere
24 GB
$1.01/hr
On-demand
AAWS
8
](https://fastgpu.co/gpus/a10)[
A30
JLJarvis Labs
NVIDIA · Ampere
24 GB
$0.41/hr
On-demand
JLJarvis Labs
1
](https://fastgpu.co/gpus/a30)[
RTX PRO 4000
RRunPod
NVIDIA · Blackwell
24 GB
$0.57/hr
On-demand
RRunPod
1
](https://fastgpu.co/gpus/rtx-pro-4000)[
RTX 4000 Ada
RRunPod
NVIDIA · Ada Lovelace
20 GB
$0.20/hr
Community
RRunPod
3
](https://fastgpu.co/gpus/rtx-4000-ada)[
P100
VAVast.ai
NVIDIA · Pascal
16 GB
$0.10/hr
On-demandLow stock
VAVast.ai
5
](https://fastgpu.co/gpus/p100)[
V100
VVerda
NVIDIA · Volta
16 GB
$0.17/hr
On-demand
VVerda
11
](https://fastgpu.co/gpus/v100)[
RTX 5090
VAVast.ai
NVIDIA · Blackwell
32 GB
$0.21/hr
On-demandIn stock
VAVast.ai
4
](https://fastgpu.co/gpus/rtx-5090)[
RTX 3090
VAVast.ai
NVIDIA · Ampere
24 GB
$0.07/hr
On-demandIn stock
VAVast.ai
4
](https://fastgpu.co/gpus/rtx-3090)[
RTX 4090
VAVast.ai
NVIDIA · Ada Lovelace
24 GB
$0.13/hr
On-demandIn stock
VAVast.ai
5
](https://fastgpu.co/gpus/rtx-4090)[
RTX A4000
VAVast.ai
NVIDIA · Ampere
16 GB
$0.07/hr
On-demandIn stock
VAVast.ai
5
](https://fastgpu.co/gpus/rtx-a4000)[
RTX 4080
RRunPod
NVIDIA · Ada Lovelace
16 GB
$0.27/hr
Community
RRunPod
1
](https://fastgpu.co/gpus/rtx-4080)[
RTX 5080
RRunPod
NVIDIA · Blackwell
16 GB
$0.39/hr
Community
RRunPod
1
](https://fastgpu.co/gpus/rtx-5080)[
RTX 3080
RRunPod
NVIDIA · Ampere
10 GB
$0.17/hr
Community
RRunPod
1
](https://fastgpu.co/gpus/rtx-3080)[
T4
GCGoogle Cloud
NVIDIA · Turing
16 GB
$0.36/hr
On-demand
GCGoogle Cloud
9
Biggest movers this week
[A100 40GB
$0.14/hr → $1.00/hr+635.3%
](https://fastgpu.co/gpus/a100-40gb)[MI300X
$1.71/hr → $2.39/hr+39.8%
](https://fastgpu.co/gpus/mi300x)[H100 SXM
$1.34/hr → $1.99/hr+49%
](https://fastgpu.co/gpus/h100-sxm)[H200
$2.60/hr → $1.98/hr-24%
](https://fastgpu.co/gpus/h200)[A100 80GB
$0.59/hr → $0.89/hr+50.1%
](https://fastgpu.co/gpus/a100-80gb)[L4
$0.20/hr → $0.49/hr+142.6%
](https://fastgpu.co/gpus/l4)[RTX PRO 6000
$0.67/hr → $0.87/hr+30%
](https://fastgpu.co/gpus/rtx-pro-6000)[H100 NVL
$2.40/hr → $2.59/hr
](https://fastgpu.co/gpus/h100-nvl)
Not a marketplace. The whole market.
FastGPU doesn’t sell GPUs. It reads the live rate for every card from every provider and shows you the cheapest, so the number you see is the real one, not a placement someone paid for.
providers, one feed
Marketplaces, neoclouds and hyperscalers normalized into a single, cheapest-first comparison.
GPU models tracked
From prosumer cards to B200 and MI300X, every model priced across every provider that lists it.
Live
every price stamped
Prices come straight from the sources on a tight refresh cycle, each row stamped with when it was fetched.
Rankings are live price first, and any paid placement is labeled. How we rank.
Cheapest GPU for your workload
Pick the job you have and see the cheapest live GPU that fits it, ranked across every provider: serving, fine-tuning, training, image generation and more.
LLM inferenceLlama 3 70BDeepSeek V3Fine-tuning LLMsTraining / full fine-tuneStable DiffusionFluxWhisper transcriptionAll workloads →
Questions
What is the cheapest cloud GPU?
The cheapest cloud GPU depends on the card and the pricing mode. Marketplaces like Vast.ai and RunPod Community list the lowest rates, often several times below hyperscalers. FastGPU ranks the live cheapest option for every GPU across the whole market.
How do I compare GPU prices across providers?
FastGPU normalizes on-demand, community and spot prices from marketplaces, neoclouds and hyperscalers into one cheapest-first table per GPU, so you can see the real spread for any card (H100, H200, B200, A100, MI300X and more) at a glance, each row stamped with when it was fetched.
Is renting a GPU cheaper than a hyperscaler?
Usually. Neoclouds and marketplaces run several times cheaper than AWS, Azure, GCP or Oracle for the same silicon. FastGPU shows all of them side by side, cheapest-first, so you can see the live spread for any card.
Live data from RunPod, Vast.ai, Verda, and the open gpuhunt catalog (AWS, Azure, GCP, Oracle, Lambda, Nebius and more). Available as a free open dataset.
Live prices·updated just now
+7.8%