> ## Documentation Index
> Fetch the complete documentation index at: https://docs.biom.science/llms.txt
> Use this file to discover all available pages before exploring further.

# Pricing

> Compute costs for running models on cloud GPUs.

# Pricing

When using Modal Labs (the default cloud provider), compute is billed per second of GPU usage. Relay automatically selects the cheapest GPU that meets the model's VRAM requirements.

## GPU pricing

| GPU           | VRAM  | Price     |
| ------------- | ----- | --------- |
| **T4**        | 16 GB | \$0.59/hr |
| **L4**        | 24 GB | \$0.80/hr |
| **A10G**      | 24 GB | \$1.10/hr |
| **A100 40GB** | 40 GB | \$2.10/hr |
| **A100 80GB** | 80 GB | \$2.50/hr |
| **H100**      | 80 GB | \$3.95/hr |
| **CPU only**  | —     | \$0.04/hr |

## Cost estimation

Before running a model, Relay estimates the compute cost:

```
Cost = (estimated_duration_seconds / 3600) × hourly_rate
```

The estimate appears in the job confirmation dialog so you know the cost before committing.

## Typical costs per model

| Model              | GPU     | Duration  | Estimated Cost  |
| ------------------ | ------- | --------- | --------------- |
| **SAM3** (image)   | T4      | \~20 sec  | \~\$0.003       |
| **SAM3** (video)   | A100    | \~60 sec  | \~\$0.035       |
| **Cellpose**       | T4      | \~30 sec  | \~\$0.005       |
| **DeepLabCut**     | T4      | \~300 sec | \~\$0.049       |
| **Suite2p**        | A10G    | \~600 sec | \~\$0.183       |
| **SpikeInterface** | T4/A10G | \~900 sec | \~$0.147–$0.275 |
| **PlantCV**        | CPU     | \~30 sec  | \~\$0.0003      |
| **SExtractor**     | CPU     | \~20 sec  | \~\$0.0002      |

## Per-user limits

| Limit                 | Default      |
| --------------------- | ------------ |
| Concurrent jobs       | 2            |
| Queued jobs           | 5            |
| Max input size        | 500 MB       |
| Max job runtime       | 3600 seconds |
| Daily compute spend   | \$5/day      |
| Monthly compute spend | \$25/month   |

These limits can be adjusted by workspace administrators.

## Other providers

* **Local Docker** — free (uses your own hardware)
* **HPC/SLURM** — costs determined by your institution
* **User GPU Server** — costs determined by your infrastructure
* **HuggingFace Spaces** — free for public spaces, paid for private