Cheapest on-demand NVIDIA L40S (48 GB) on 2026-10-10 14:06 UTC: $0.536 per GPU-hour on Vast.ai. Prices from 3 providers, checked every hour.
What can I run on it? Open the calculator →| Provider | Per GPU-hour | 1 GPU, per month | Rent |
|---|---|---|---|
| Vast.ai | $0.536 | $391.28 | Rent on Vast.ai (opens in a new tab) → |
| RunPod | $1.09 | $795.70 | Rent on RunPod (opens in a new tab) → |
| Verda | $1.62 | $1,184.06 | Rent on Verda (opens in a new tab) → |
On-demand, per GPU, checked every hour. Monthly = 730 hours. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
Loading the daily history…
With 4-bit weights and an 8K-token context, a single NVIDIA L40S runs: Ornith 1.5 35B-A3B, Qwen3.6 35B-A3B, Qwen3 32B, Nemotron 3 Nano 30B-A3B, Gemma 4 31B, Qwen3-Coder 30B-A3B, JEV 27B VL, Qwen3.8 27B, Clef, Gemma 4 26B-A4B, Mistral Small 3.2 24B, LFM2 24B-A2B, gpt-oss 20B, Mellum2.1 12B-A2.5B, Gemma 4 12B, Humanizer, Ornith 1.5 9B, Qwen3.5 9B, Clef Flash, LFM2.5 8B-A1B, Llama 3.1 8B, Qwen3.5 4B, LightOnOCR-3 4B, Spark-X2.5 4B, D1 3B.
On 2026-10-10, the cheapest on-demand NVIDIA L40S costs $0.536 per GPU-hour on Vast.ai; across 3 providers prices range from $0.536 to $1.62. An 8-GPU server costs about $4.29 per hour, and one GPU running all month about $391.28.
Vast.ai at $0.536 per hour on 2026-10-10. Marketplace prices (Vast.ai) depend on the host and move within the day; the history on this page shows how they evolve.
With 4-bit weights and an 8K-token context: Ornith 1.5 35B-A3B, Qwen3.6 35B-A3B, Qwen3 32B, Nemotron 3 Nano 30B-A3B, Gemma 4 31B, Qwen3-Coder 30B-A3B, JEV 27B VL, Qwen3.8 27B, Clef, Gemma 4 26B-A4B, Mistral Small 3.2 24B, LFM2 24B-A2B, gpt-oss 20B, Mellum2.1 12B-A2.5B, Gemma 4 12B, Humanizer, Ornith 1.5 9B, Qwen3.5 9B, Clef Flash, LFM2.5 8B-A1B, Llama 3.1 8B, Qwen3.5 4B, LightOnOCR-3 4B, Spark-X2.5 4B, D1 3B. Larger models need several GPUs; the LLM VRAM calculator sizes them.
NVIDIA B300 / GB300 · NVIDIA B200 (HGX) · NVIDIA H200 · NVIDIA H100 NVL · NVIDIA H100 · NVIDIA A100 80GB · NVIDIA A100 40GB · NVIDIA L4 · AMD MI355X / MI350X · AMD MI300X · RTX PRO 6000 Blackwell · RTX 6000 Ada · GeForce RTX 5090 · GeForce RTX 4090 · GeForce RTX 3090 · GeForce RTX 5080