H100

Memory80 GB
Clouds10
Lowest / hour$2.08Seeweb
Largest machine8 GPUs

Clouds

Lowest price first
CloudPer GPU-hourMachineSizesRegions
Seeweb$2.081 GPU1, 2, 4—
Lambda$3.791 GPU1, 2, 4, 817
Modal$3.951 GPU1, 2, 3, 4, 5, 6, 7, 823
Nebius$4.501 GPU1, 81
Verda$5.242 GPUs21
AWS$6.881 GPU1, 813
Oracle Cloud$10.008 GPUs8—
Google Cloud$10.984 GPUs1, 2, 4, 819
Microsoft Azure$12.298 GPUs823

On one H100, 80 GB

Highest index first
#ModelIndexFormatMemory
01Qwen3.8-27B58.816-bit59.3 GB
02Muse Glimmer 30B52.816-bit63.5 GB
03Ripple-65B Preview52.08-bit73.3 GB
04Ternary Bonsai 2 27B52.016-bit57.5 GB
05Qwen3.6-27B51.916-bit59.3 GB

On 8 × H100, 640 GB

Highest index first
#ModelIndexFormatMemory
01MiniMax M3.1 Flash69.016-bit464.0 GB
02DeepSeek V4.1 Flash 101768.14-bit486.8 GB
03Step 5 Preview68.04-bit385.7 GB
04Hy4 preview66.44-bit497.5 GB
05GLM-5.365.64-bit480.5 GB

The lowest on-demand list price per GPU-hour in each cloud, in US dollars, from the SkyPilot catalog, the Oracle Cloud price list and the Azure Retail Prices API, read 16 Nov 2026. Machine: the GPUs in the machine with that price. The models are ranked open models whose weights, with room for a short context, fit in the memory. Prices change often: check the cloud before you rent.