GPU class · 80 GB
H100 80 GB
What does H100 80 GB cost and what fits on it?
H100 80 GB has 80 GB and costs $2.59 per hour to train on, $2.99 to serve. 40 models in the catalogue fit half-precision LoRA on it and 3 more fit in four-bit.
Same capacity as an A100 80GB, substantially more throughput, and fp8 support. On a long run the wall-time saving can pay for the rate difference; on a short one it will not. Worth being deliberate about, because this is the class where an idle endpoint gets expensive fastest.
When to pick this class
Same capacity as an A100 80GB, substantially more throughput, and fp8 support. On a long training run the wall-time saving can pay for the rate difference outright; on a short one it will not, and the arithmetic is worth doing rather than assuming the faster card is the professional choice.
This is also the class where an idle endpoint gets expensive fastest. A continuously warm replica here costs several times what the same replica costs on a 48GB class, so be deliberate about the scaling mode rather than defaulting to always-warm.
Rates
| VRAM | 80 GB |
|---|---|
| Training | $2.59 / GPU-hourMetered per GPU-second, one-minute floor |
| Serving | $2.99 / GPU-hourWhile a replica is resident |
| Warm for a day | $71.7624 hours resident, regardless of traffic |
| Warm for 30 days | $2152.80Set the endpoint to scale to zero if nobody is waiting |
| Models — LoRA | 40 |
| Models — QLoRA only | 3 |
Largest models that fit for LoRA
Half precision, frozen base, within 80 GB.
Qwen3 32B
32BNeeds 64 GB · Apache 2.0
OLMo 3 32B
32BNeeds 64 GB · Apache 2.0
Qwen3 30B-A3B
30BNeeds 64 GB · Apache 2.0
Gemma 3 27B
27BNeeds 56 GB · Gemma Terms of Use
Mistral Small 3 24B
24BNeeds 48 GB · Apache 2.0
LFM2 24B-A2B
24BNeeds 48 GB · LFM Open
gpt-oss 20B
20.9BNeeds 40 GB · Apache 2.0
Qwen3 14B
14BNeeds 28 GB · Apache 2.0
Models that need four-bit training here
These exceed 80 GB in half precision but fit quantised. Quantised training is slower per step, so compare the total run cost against the next class up rather than the hourly rate alone.
Compared with its neighbours
The classes immediately either side of H100 80 GB on capacity and rate.
| Class | VRAM | Training | Serving | Versus this one |
|---|---|---|---|---|
| L40S | 48 GB | $0.78 | $0.90 | 32 GB less, $1.81/hr cheaper |
| A100 80 GB | 80 GB | $1.68 | $1.94 | same capacity, $0.91/hr cheaper |
| H200 | 141 GB | $4.55 | $5.25 | 61 GB more, $1.96/hr dearer |
Frequently asked questions
How much does H100 80 GB cost per hour?
$2.59 per GPU-hour for training and $2.99 for serving, metered per GPU-second above a one-minute floor. A continuously warm endpoint on this class is about $71.76 a day, or $2152.80 over thirty days.
What models fit on H100 80 GB?
40 models in the catalogue fit half-precision LoRA training within 80 GB, and 3 more fit in four-bit. 42 can be served from this class before accounting for the attention cache.
Is H100 80 GB the cheapest option?
Cheapest per hour is not the same as cheapest per run. Total cost is the rate multiplied by wall time, so a faster class that finishes sooner often costs less — particularly when the cheaper alternative would require quantised training, which is slower per step.
Last verified 6 August 2026.
Start with the free tier
A magic link creates your account, your tenant and your first API key. No card until you ask for compute.