Hardware Pricing
Detailed pricing for all Skytells hardware tiers — CPUs, GPUs, and multi-GPU configurations.
Hardware Pricing
Models running on Skytells infrastructure are billed per second based on the hardware they use. You only pay for the time your request takes to process.
Pricing shown is per second of compute time. You are not charged while the model is idle — only during active processing of your request.
Standard Hardware
| Hardware | ID | Price/sec | Price/hr | GPU | CPU | GPU RAM | RAM |
|---|---|---|---|---|---|---|---|
| CPU (Small) | cpu-small | $0.000025 | $0.09 | — | 1x | — | 2 GB |
| CPU | cpu | $0.000100 | $0.36 | — | 4x | — | 8 GB |
| Nvidia T4 GPU | gpu-t4 | $0.001225 | $4.41 | 1x | 4x | 16 GB | 16 GB |
| Nvidia L40S GPU | gpu-l40s | $0.001975 | $7.11 | 1x | 10x | 48 GB | 65 GB |
| 2x Nvidia L40S GPU | gpu-l40s-2x | $0.002950 | $10.62 | 2x | 20x | 96 GB | 144 GB |
| Nvidia A100 (80 GB) GPU | gpu-a100-large | $0.002400 | $8.64 | 1x | 10x | 80 GB | 144 GB |
| 2x Nvidia A100 (80 GB) GPU | gpu-a100-large-2x | $0.003800 | $13.68 | 2x | 20x | 160 GB | 288 GB |
| Nvidia H100 GPU | gpu-h100 | $0.002525 | $9.09 | 1x | 13x | 80 GB | 72 GB |
Additional Multi-GPU Hardware
The following multi-GPU configurations are available with committed spend contracts. Contact Support for details.
| Hardware | ID | Price/sec | Price/hr |
|---|---|---|---|
| 4x Nvidia A100 (80 GB) GPU | gpu-a100-large-4x | $0.006600 | $23.76 |
| 8x Nvidia A100 (80 GB) GPU | gpu-a100-large-8x | $0.012200 | $43.92 |
| 2x Nvidia H100 GPU | gpu-h100-2x | $0.004050 | $14.58 |
| 4x Nvidia H100 GPU | gpu-h100-4x | $0.007100 | $25.56 |
| 8x Nvidia H100 GPU | gpu-h100-8x | $0.013200 | $47.52 |
| 4x Nvidia L40S GPU | gpu-l40s-4x | $0.004900 | $17.64 |
| 8x Nvidia L40S GPU | gpu-l40s-8x | $0.008800 | $31.68 |
Multi-GPU configurations beyond standard tiers require a committed spend contract. Reach out to our team for custom pricing and availability.
How Hardware Billing Works
- You submit a request — your prompt or input is sent to the model.
- Hardware is allocated — the model runs on the appropriate hardware tier.
- You're billed for processing time — only the seconds spent actively processing your request are billed. Idle time is not charged.
The hardware tier used by a model is shown on the model's page in Dashboard → Console.
Cost Estimation
You'll find cost estimates for any model on its page. The estimate is based on the hardware tier and typical processing time for that model.
For workloads with predictable usage, consider a committed spend contract for discounted rates on multi-GPU hardware.
Need Help?
Contact Support
Submit a support ticket for hardware pricing or committed spend inquiries.
How is this guide?