Calyx charges a 15% platform fee on every completed job. Hosts keep 85%. Renters pay only for what they use, metered by the second.
| GPU model | VRAM | Renter pays / hr | Host earns / hr | Best for |
|---|---|---|---|---|
| RTX 3090 | 24 GB | $0.22 | $0.19 | Inference, fine-tuning |
| RTX 4090 Popular | 24 GB | $0.34 | $0.29 | Inference, small training |
| A100 40GB | 40 GB | $0.65 | $0.55 | Model training |
| A100 80GB | 80 GB | $1.20 | $1.02 | Large model training |
| H100 PCIe High demand | 80 GB | $2.10 | $1.79 | LLM training, inference |
| H100 SXM | 80 GB | $3.50 | $2.98 | Frontier model training |
Rates are benchmarked against Vast.ai and RunPod as of Q1 2026. Hosts may set custom rates within ±20% of the above.
List your GPU in under 5 minutes. First 100 hosts pay zero platform fee for 6 months.