RRunyard Index
Runyard / GPU Index / Compare

Compare

Two GPUs, or two providers, side by side — from the same index, priced 2026-09-18. GPU pages compare rental price, memory, bandwidth, which models fit on one card and what a million generated tokens costs on each. Provider pages compare the two on every GPU both sell.

GPU vs GPU

H100 SXM vs A100 80GB$3.50 vs $1.79 per GPU-hour · 80 vs 80 GBH100 SXM vs H200$3.50 vs $4.45 per GPU-hour · 80 vs 141 GBH200 vs B200$4.45 vs $6.59 per GPU-hour · 141 vs 180 GBH100 SXM vs B200$3.50 vs $6.59 per GPU-hour · 80 vs 180 GBB200 vs B300$6.59 vs $7.85 per GPU-hour · 180 vs 288 GBH100 SXM vs H100 PCIe$3.50 vs $2.88 per GPU-hour · 80 vs 80 GBH100 SXM vs H100 NVL$3.50 vs $3.03 per GPU-hour · 80 vs 94 GBH100 SXM vs MI300X$3.50 vs $2.59 per GPU-hour · 80 vs 192 GBH200 vs MI300X$4.45 vs $2.59 per GPU-hour · 141 vs 192 GBH200 vs MI325X$4.45 vs $3.20 per GPU-hour · 141 vs 256 GBB200 vs MI355X$6.59 vs $8.60 per GPU-hour · 180 vs 288 GBA100 80GB vs A100 40GB$1.79 vs $1.48 per GPU-hour · 80 vs 40 GBA100 80GB vs L40S$1.79 vs $1.47 per GPU-hour · 80 vs 48 GBL40S vs L4$1.47 vs $0.80 per GPU-hour · 48 vs 24 GBL40S vs RTX 6000 Ada$1.47 vs $0.82 per GPU-hour · 48 vs 48 GBL40S vs RTX PRO 6000 Blackwell$1.47 vs $2.05 per GPU-hour · 48 vs 96 GBA100 80GB vs RTX PRO 6000 Blackwell$1.79 vs $2.05 per GPU-hour · 80 vs 96 GBH100 SXM vs RTX PRO 6000 Blackwell$3.50 vs $2.05 per GPU-hour · 80 vs 96 GBGH200 vs H100 SXM$3.02 vs $3.50 per GPU-hour · 96 vs 80 GBH200 vs GH200$4.45 vs $3.02 per GPU-hour · 141 vs 96 GBRTX PRO 6000 Blackwell vs RTX 6000 Ada$2.05 vs $0.82 per GPU-hour · 96 vs 48 GBRTX 6000 Ada vs RTX A6000$0.82 vs $0.56 per GPU-hour · 48 vs 48 GBRTX A6000 vs RTX A5000$0.56 vs $0.27 per GPU-hour · 48 vs 24 GBRTX A5000 vs RTX A4000$0.27 vs $0.17 per GPU-hour · 24 vs 16 GBRTX 5090 vs RTX 4090$0.69 vs $0.60 per GPU-hour · 32 vs 24 GBRTX 4090 vs RTX 3090$0.60 vs $0.25 per GPU-hour · 24 vs 24 GBRTX 5090 vs RTX 5080$0.69 vs $0.53 per GPU-hour · 32 vs 16 GBRTX 4090 vs RTX 4080$0.60 vs $0.31 per GPU-hour · 24 vs 16 GBRTX 5080 vs RTX 4080$0.53 vs $0.31 per GPU-hour · 16 vs 16 GBRTX 4090 vs A100 80GB$0.60 vs $1.79 per GPU-hour · 24 vs 80 GBRTX 5090 vs A100 80GB$0.69 vs $1.79 per GPU-hour · 32 vs 80 GBRTX 4090 vs L40S$0.60 vs $1.47 per GPU-hour · 24 vs 48 GBRTX 4090 vs RTX A6000$0.60 vs $0.56 per GPU-hour · 24 vs 48 GBRTX 5090 vs RTX PRO 6000 Blackwell$0.69 vs $2.05 per GPU-hour · 32 vs 96 GBA10 vs A10G$1.29 vs $1.83 per GPU-hour · 24 vs 24 GBA10 vs L4$1.29 vs $0.80 per GPU-hour · 24 vs 24 GBT4 vs L4$0.59 vs $0.80 per GPU-hour · 16 vs 24 GBV100 32GB vs A100 40GB$2.66 vs $1.48 per GPU-hour · 32 vs 40 GBV100 16GB vs T4$0.79 vs $0.59 per GPU-hour · 16 vs 16 GBA40 vs RTX A6000$0.65 vs $0.56 per GPU-hour · 48 vs 48 GB

Provider vs provider

Every pair among the widest-coverage providers that share at least three GPUs.

RunPod Community Cloud vs RunPod Secure Cloud23 GPUs in commonRunPod Community Cloud vs Vast.ai23 GPUs in commonRunPod Community Cloud vs GPU.ai21 GPUs in commonRunPod Community Cloud vs io.net19 GPUs in commonRunPod Community Cloud vs Runcrate14 GPUs in commonRunPod Community Cloud vs Massed Compute12 GPUs in commonRunPod Community Cloud vs Oracle Cloud10 GPUs in commonRunPod Community Cloud vs Spheron11 GPUs in commonRunPod Community Cloud vs Akash10 GPUs in commonRunPod Community Cloud vs Amazon Web Services8 GPUs in commonRunPod Community Cloud vs Modal9 GPUs in commonRunPod Community Cloud vs Beam10 GPUs in commonRunPod Community Cloud vs Cerebrium8 GPUs in commonRunPod Secure Cloud vs Vast.ai22 GPUs in commonRunPod Secure Cloud vs GPU.ai20 GPUs in commonRunPod Secure Cloud vs io.net18 GPUs in commonRunPod Secure Cloud vs Runcrate12 GPUs in commonRunPod Secure Cloud vs Massed Compute12 GPUs in commonRunPod Secure Cloud vs Oracle Cloud8 GPUs in commonRunPod Secure Cloud vs Spheron11 GPUs in commonRunPod Secure Cloud vs Akash9 GPUs in commonRunPod Secure Cloud vs Amazon Web Services6 GPUs in commonRunPod Secure Cloud vs Modal8 GPUs in commonRunPod Secure Cloud vs Beam10 GPUs in commonRunPod Secure Cloud vs Cerebrium7 GPUs in commonVast.ai vs GPU.ai20 GPUs in commonVast.ai vs io.net18 GPUs in commonVast.ai vs Runcrate13 GPUs in commonVast.ai vs Massed Compute12 GPUs in commonVast.ai vs Oracle Cloud8 GPUs in commonVast.ai vs Spheron10 GPUs in commonVast.ai vs Akash10 GPUs in commonVast.ai vs Amazon Web Services7 GPUs in commonVast.ai vs Modal8 GPUs in commonVast.ai vs Beam10 GPUs in commonVast.ai vs Cerebrium7 GPUs in commonGPU.ai vs io.net17 GPUs in commonGPU.ai vs Runcrate12 GPUs in commonGPU.ai vs Massed Compute11 GPUs in commonGPU.ai vs Oracle Cloud7 GPUs in commonGPU.ai vs Spheron9 GPUs in commonGPU.ai vs Akash8 GPUs in commonGPU.ai vs Amazon Web Services7 GPUs in commonGPU.ai vs Modal8 GPUs in commonGPU.ai vs Beam10 GPUs in commonGPU.ai vs Cerebrium8 GPUs in commonio.net vs Runcrate14 GPUs in commonio.net vs Massed Compute10 GPUs in commonio.net vs Oracle Cloud8 GPUs in commonio.net vs Spheron10 GPUs in commonio.net vs Akash9 GPUs in commonio.net vs Amazon Web Services9 GPUs in commonio.net vs Modal10 GPUs in commonio.net vs Beam10 GPUs in commonio.net vs Cerebrium9 GPUs in commonRuncrate vs Massed Compute8 GPUs in commonRuncrate vs Oracle Cloud8 GPUs in commonRuncrate vs Spheron7 GPUs in commonRuncrate vs Akash5 GPUs in commonRuncrate vs Amazon Web Services8 GPUs in commonRuncrate vs Modal8 GPUs in commonRuncrate vs Beam7 GPUs in commonRuncrate vs Cerebrium8 GPUs in commonMassed Compute vs Oracle Cloud6 GPUs in commonMassed Compute vs Spheron7 GPUs in commonMassed Compute vs Akash4 GPUs in commonMassed Compute vs Amazon Web Services4 GPUs in commonMassed Compute vs Modal6 GPUs in commonMassed Compute vs Beam6 GPUs in commonMassed Compute vs Cerebrium5 GPUs in commonOracle Cloud vs Spheron8 GPUs in commonOracle Cloud vs Akash6 GPUs in commonOracle Cloud vs Amazon Web Services7 GPUs in commonOracle Cloud vs Modal9 GPUs in commonOracle Cloud vs Beam6 GPUs in commonOracle Cloud vs Cerebrium8 GPUs in commonSpheron vs Akash7 GPUs in commonSpheron vs Amazon Web Services5 GPUs in commonSpheron vs Modal7 GPUs in commonSpheron vs Beam8 GPUs in commonSpheron vs Cerebrium6 GPUs in commonAkash vs Amazon Web Services6 GPUs in commonAkash vs Modal6 GPUs in commonAkash vs Beam6 GPUs in commonAkash vs Cerebrium5 GPUs in commonAmazon Web Services vs Modal8 GPUs in commonAmazon Web Services vs Beam5 GPUs in commonAmazon Web Services vs Cerebrium8 GPUs in commonModal vs Beam6 GPUs in commonModal vs Cerebrium10 GPUs in commonBeam vs Cerebrium6 GPUs in common

Questions

H100 vs A100 — which should I rent?

On 2026-09-18 the H100 SXM median is $3.50 per GPU-hour and the A100 80GB $1.79. Both hold 80 GB, so the same models fit on one card; the H100's memory bandwidth is 3,350 GB/s against 2,039, so it generates tokens roughly 1.6× faster per card. If the H100 costs less than 1.6× the A100 where you rent, it is the better per-token deal; the comparison page prices this per model.

H200 vs B200 — is Blackwell worth it?

H200 median $4.45, B200 $6.59. The B200 has 180 GB against 141 and 8,000 GB/s against 4,800, so it holds larger models on one card and streams them faster. For a model that already fits on an H200, the question is whether the B200's bandwidth premium is smaller than its price premium — the page computes cost per million tokens on each.

RTX 4090 vs RTX 5090 for renting?

RTX 4090 median $0.60, RTX 5090 $0.69. The 5090's 32 GB against 24 lets a 4-bit 30B model with room for context fit where the 4090 is tight, and its 1,792 GB/s against 1,008 is a large speed step. Both are consumer cards mostly on marketplaces, so availability and host quality vary.

How do the provider comparisons work?

Every pair among the widest-coverage providers that share at least three GPUs gets a page: each GPU both sell, with both prices, who is cheaper on each, and the market median beside them. Only GPUs both actually price appear, so it is a like-for-like table, not a marketing sheet. Pairs are rebuilt from each snapshot.

What does "cost per million tokens" mean on these pages?

The GPU's hourly median divided by an estimate of how many tokens it generates in an hour for that model, single stream, from memory bandwidth — every token streams the active weights once. It is a floor, not a benchmark, and batching lowers it by an order of magnitude; it exists to rank cards against each other for the same model.