GPU class · 24 GB
RTX 4090
What does RTX 4090 cost and what fits on it?
RTX 4090 has 24 GB and costs $0.38 per hour to train on, $0.43 to serve. 31 models in the catalogue fit half-precision LoRA on it and 3 more fit in four-bit.
The class most runs on this platform end up using. Twenty-four gigabytes covers LoRA on the whole 4B to 8B band — where the great majority of production fine-tunes live — and it is fast enough that the shorter wall time usually offsets the higher hourly rate against a 3090. It also serves an 8B base with several adapters resident at once.
When to pick this class
More runs on this platform end up here than anywhere else, and the reason is the 4B-to-8B band. That is where most production fine-tunes live, half-precision LoRA fits in twenty-four gigabytes across the whole band, and the throughput is high enough that the shorter wall time usually offsets the higher hourly rate against a 3090.
It is also a capable serving class for an 8B base with several adapters resident at once. If you are running a multi-variant deployment on a mainstream budget, this is the class to size against — and the marginal cost of each additional adapter is megabytes rather than gigabytes.
Rates
| VRAM | 24 GB |
|---|---|
| Training | $0.38 / GPU-hourMetered per GPU-second, one-minute floor |
| Serving | $0.43 / GPU-hourWhile a replica is resident |
| Warm for a day | $10.3224 hours resident, regardless of traffic |
| Warm for 30 days | $309.60Set the endpoint to scale to zero if nobody is waiting |
| Models — LoRA | 31 |
| Models — QLoRA only | 3 |
Largest models that fit for LoRA
Half precision, frozen base, within 24 GB.
Mistral Nemo 12B
12BNeeds 24 GB · Apache 2.0
Gemma 3 12B
12BNeeds 24 GB · Gemma Terms of Use
Falcon 3 10B
10BNeeds 22 GB · TII Falcon LLM
LFM2 8B-A1B
8.3BNeeds 18 GB · LFM Open
Qwen3 8B
8BNeeds 18 GB · Apache 2.0
Llama 3.1 8B
8BNeeds 18 GB · Llama 3.1 Community
DeepSeek-R1 Distill Llama 8B
8BNeeds 18 GB · Llama 3.1 Community
Granite 3.0 8B
8BNeeds 18 GB · Apache 2.0
Models that need four-bit training here
These exceed 24 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 RTX 4090 on capacity and rate.
| Class | VRAM | Training | Serving | Versus this one |
|---|---|---|---|---|
| L4 | 24 GB | $0.17 | $0.20 | same capacity, $0.21/hr cheaper |
| RTX 3090 | 24 GB | $0.22 | $0.26 | same capacity, $0.16/hr cheaper |
| A100 40 GB | 40 GB | $1.17 | $1.35 | 16 GB more, $0.79/hr dearer |
| A40 | 48 GB | $0.42 | $0.48 | 24 GB more, $0.04/hr dearer |
Frequently asked questions
How much does RTX 4090 cost per hour?
$0.38 per GPU-hour for training and $0.43 for serving, metered per GPU-second above a one-minute floor. A continuously warm endpoint on this class is about $10.32 a day, or $309.60 over thirty days.
What models fit on RTX 4090?
31 models in the catalogue fit half-precision LoRA training within 24 GB, and 3 more fit in four-bit. 32 can be served from this class before accounting for the attention cache.
Is RTX 4090 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.