NVIDIA L4
Ada Lovelace Architecture
Active
Launched March 2023
Core Specifications
VendorNVIDIA
ArchitectureAda Lovelace
Form FactorPCIe
VRAM24 GB
Memory Bandwidth300 GB/s
TDP72 W
Compute Performance
| Precision | TFLOPs |
|---|---|
| FP32 | 30.3 |
| FP16 | 242 |
| BF16 | 121 |
| FP8 | 485 |
Performance Benchmarks
llm inference
| Configuration | Precision | Performance | Source |
|---|---|---|---|
| 1x | FP8 | 1,800 tokens/sec | View |
Pricing
Hardware Purchase (CAPEX)
| Type | Price (USD) | Region | As of |
|---|---|---|---|
| Street Price | $3,500 | Global | Oct 2024 |
| Street Price | $4,000 | global | Oct 2025 |
Cloud Rental (OPEX)
| Provider | Instance Type | Price per Hour | Region | As of |
|---|---|---|---|---|
| OVH | — | $1.00/hr | global | Nov 2025 |
| Google Cloud | — | $1.00/hr | global | Nov 2025 |
| Sesterce | — | $0.99/hr | global | Nov 2025 |
| AWS | — | $0.98/hr | global | Nov 2025 |
| Scaleway | — | $0.90/hr | global | Nov 2025 |
| Google Cloud | — | $0.85/hr | global | Nov 2025 |
| AWS | — | $0.80/hr | global | Nov 2025 |
| Google Cloud | — | $0.71/hr | global | Nov 2025 |
| Koyeb | — | $0.70/hr | global | Nov 2025 |
| Runpod | — | $0.44/hr | global | Nov 2025 |
| RunPod | 1x L4 (24GB) | $0.44/hr | Global | Oct 2024 |
| AWS | g6.xlarge (1x L4) | $1.10/hr | us-east-1 | Oct 2024 |
| Google Cloud | g2-standard-4 (1x L4) | $0.80/hr | us-central1 | Oct 2024 |
| Google Cloud | — | $0.80/hr | us-east | Oct 2025 |