⚠️ Use prices only as an estimate—check the pricing pages on the official website!
| Provider | Services | Custom LLM hosting | Custom LLM pricing | Shared models | Shared model pricing |
|---|---|---|---|---|---|
| beam.cloud | Training + Inference | By the second | $3.29/h (A100 40GB) | No | No |
| banana.dev | Inference | By the second | $7.49/h (A100 40GB) | No | No |
| fal.ai | Inference | By the second | $7.02/h (A100 40GB) | Lots | By the second |
| together.ai | Training + Inference | By the second | Based on model size | Lots | Per 1K tokens |
| replicate.com | Inference | By the second | $8.28/h (A100 40GB) | Lots | $4.14/h |
| brev.dev | Training + Inference | Hourly (?) | $1.10-3.67 (A100 40GB) | ? | ? |
| gradient.ai | Training + Inference | Per 1K tokens | Based on model size | ||
| titanml.co | Training + Inference | ? | ? | ? | ? |
| openpipe.ai | Training + Inference | Per 1K tokens | Based on model size | Same as custom | Same as custom |
| endpoints.anyscale.com | Inference | No | No | Llama 2, Code Llama | Per 1M tokens |
| endpoints.huggingface.co | |||||
| modal.com | |||||
| fireworks.ai | |||||
| lepton.ai |
For all options, latency and throughput are unknown factors.