STRATA Reserve capacity
Strata/Answers/Guides

Which GPU to rent for models needing 80 GB, 130 GB or more VRAM

GuidesUpdated 2026-10-23

Match the GPU to total memory need, weights plus 20–50% for KV cache: up to about 80 GB fits one H100 ($3.23/hour at Strata GPU) or one 96 GB RTX PRO 6000 ($2.03); 100–140 GB needs one H200 141GB ($4.31) or two H100s; above 141 GB use 2–8 H200s, or 8-GPU nodes with 640 GB (H100), 1,128 GB (H200) or about 1,440 GB (B200).

Reserve capacityAll GPUs and prices

Memory tiers you can rent

Memory you needSetupTotal GPU memoryStrata on demand
Up to ~80 GB1× H100 80GB80 GB$3.23/hr
Up to ~96 GB, single card1× RTX PRO 600096 GB$2.03/hr
~100–140 GB1× H200141 GB$4.31/hr
~140–160 GB2× H100 (NVLink)160 GB2 × $3.23/hr
~160–280 GB2× H200282 GB2 × $4.31/hr
Up to ~640 GB8× H100 node640 GB$18,863/month
Up to ~1.1 TB8× H200 node1,128 GB$25,170/month
Up to ~1.4 TB8× B200 node≈ 1,440 GB usableReserved, on request

Strata list prices per GPU-hour; node per month = price × 8 GPUs × 730 hours. Prices as of 2026-11-10.

How to work out how much VRAM you need

One big GPU or several smaller ones

Very large models

Models of several hundred billion parameters need a full 8-GPU node. DeepSeek V3 and R1 (671B, about 700 GB in FP8) run on one 8×H200 or 8×B200 node; on 8×H100 they need 4-bit quantization or two nodes. B200 is allocated as full 8-GPU nodes on 3–36 month reserved terms. See how many GPUs DeepSeek needs.

Sources

FAQ
Can a 130 GB model run on one H200?

Only just. The H200 has 141 GB, so 130 GB of weights leave about 11 GB for KV cache, which limits context length and concurrency. Two H100s (160 GB) or two H200s are the safer choice, or quantize to 8-bit and use one H200.

Is the RTX PRO 6000 a cheaper alternative to the H100 for 80–96 GB models?

For single-GPU inference and image or video generation, often yes: it has 96 GB and costs $2.03 per hour against $3.23 for an H100. For training and multi-GPU serving the H100 wins on memory bandwidth and NVLink.

Can I rent multi-GPU setups by the month?

Yes. On-demand GPUs are billed per minute with no term; dedicated 8-GPU nodes can be reserved for 3–36 months at 3–14% below the on-demand rate.

NVIDIA H100 80GB
Hopper · 80 GB HBM3

H100

from $3.23/GPU-hr

Specs and pricing →
NVIDIA H200 141GB
Hopper · 141 GB HBM3e

H200

from $4.31/GPU-hr

Specs and pricing →
NVIDIA RTX PRO 6000 Blackwell 96GB
Blackwell · 96 GB GDDR7

RTX PRO 6000

from $2.03/GPU-hr

Specs and pricing →
NVIDIA B200 192GB
Blackwell · 192 GB HBM3e

B200

Pricing on request

Specs and pricing →
Related questions

How much does it cost to rent an H100 per hour in 2026? How much does it cost to rent an H200 GPU? How much does it cost to rent an NVIDIA B200? Where can I rent an RTX 4090 or RTX 5090 GPU?