3 providers · per GPU-hour
NVIDIA L4 cloud pricing
3 offers from 3 providers, 3 with a published on-demand price. Per GPU-hour, US$. Prices checked 16 Sept 2026.
- Lowest on-demand
- $0.44/GPU-hr
- Lowest spot
- —
- Memory
- 24 GB
Spec sheet: L4 ↗
| # | Provider | Offer | On-demand | Spot | Committed | As published (original currency) · source | Actions |
|---|---|---|---|---|---|---|---|
| 1 | Jarvislabs | L4 x1 1× L4 · 32 vCPU · 124 GB RAM | $0.44/GPU-hr | — | — | ||
| 2 | UpCloud | L4 x1 1× L4 · 8 vCPU · 64 GB RAM | $0.65/GPU-hr | — | — | ||
| 3 | OVHcloud | L4 x1 1× L4 · 22 vCPU · 90 GB RAM · 400 GB disk | $0.73/GPU-hr | — | — |
What is the L4 used for?
The L4 is a data-centre GPU positioned by its maker for inference, built on the Ada Lovelace architecture and released in 2023. Multi-GPU servers link cards over None listed (PCIe only), which matters when a job is split across several GPUs.
How large a model fits on one L4?
With 24 GB of memory, one card holds the weights of a model up to about 12 billion parameters at 16-bit precision, or about 24 billion at 8-bit. That is weights only: inference also needs room for the KV cache, and training needs several times more for gradients and optimiser state, so larger jobs are split across multiple GPUs.
| Spec (as published) | L4 |
|---|---|
| Architecture | Ada Lovelace |
| Form factor | PCIe 1-slot low-profile |
| Memory | 24 GB |
| Memory bandwidth | 300 GB/s |
| Interconnect | None listed (PCIe only) |
| Max power (TDP) | 72 W |
| Precisions listed | FP32, TF32, FP16, BF16, FP8, INT8 |
| Released | 2023 |
| Source | maker spec sheet ↗ |
Need L4 capacity for a longer run?
Committed and cluster pricing is often negotiated. Tell us the GPU count, duration and region.