The NVIDIA L40S is the Ada Lovelace inference GPU, 48 GB of GDDR6 memory with Ada's 4th-gen Tensor Cores, delivering 733 TFLOPS of FP16 compute at a fraction of H100 cost. Ideal for cost-efficient LLM inference and generative AI. Available on packet.ai from $0.92/GPU-hour.
The L40S brings Ada's 4th-gen Tensor Cores, FP8 precision, and AV1 hardware encode to inference-focused server workloads, at GDDR6 price points.
Ada Tensor Cores with FP8 deliver up to 733 TFLOPS of FP16 compute. Strong throughput at a fraction of H100 cost.
48 GB fits 7B to 13B models natively; 30B+ at 4-bit.
Hardware AV1 encoding and 3rd-gen RT Cores make the L40S the best GPU for AI video generation.
Single-slot PCIe drops into any server without SXM motherboards.
7B–13B fit natively; 30B+ at 4-bit. High FP16 throughput at under half H100 cost.
AV1 hardware encode and Ada RT Cores make L40S ideal for SDXL, FLUX, and video generation.
LoRA and QLoRA fine-tuning of 7B–13B models at the lowest cost per experiment.
| Configuration | On-Demand | Monthly | 3 Months | 6 Months | Annually |
|---|---|---|---|---|---|
| Dedicated | Dedicated only | ||||
| 1× NVIDIA L40SMost Popular | $0.92/hr | $604/mo $0.83/hr eff. | $1,812/3mo | $3,624/6mo | $7,248/yr |
| 2× NVIDIA L40SDedicated only | $1.84/hr | $1,142/mo $1.56/hr eff. | $3,426/3mo | $6,852/6mo | $13,704/yr |
| 4× NVIDIA L40SDedicated only | $3.68/hr | $2,283/mo $3.13/hr eff. | $6,849/3mo | $13,698/6mo | $27,396/yr |
| Deploy hourly → | Subscribe to Monthly → |
The L40S is NVIDIA's Ada Lovelace inference GPU, 48 GB GDDR6, 733 TFLOPS FP16, AV1 encode, and FP8 Tensor Cores.
L40S starts at $0.92/GPU-hour dedicated. Monthly from $604/mo.
H100 has HBM3 memory (3.35 TB/s vs 864 GB/s) and higher sustained throughput. L40S is cheaper and better for small-batch inference and image generation.
7B and 13B at FP16 fit easily. 30B–70B at 4-bit. For full FP16 70B, use A100 or H100.
Yes, one of the best value GPUs for Stable Diffusion XL, FLUX, and video generation.
Production-grade inference at $0.92/hr dedicated or $604/mo flat.
On-demand · hourly billing · US & EU regions