Mosaic AI Model Training - pre-training
![](/en-website-assets/static/04df36feaaab180c807d8017a913c6a8/illustration_nodes_color_4_01666985242.png)
Loading...
Model Training - pre-training GPU Prices and DBU Rates
Instance type | DBU/hr (all regions) | On-Demand Training (US Price per GPU per hour) | Training Reservation (US Price per GPU per hour) | Training Hero Reservation (US Price per GPU per hour) |
---|---|---|---|---|
A10 | 4.361 | N/A | $2.40 | N/A |
A100 40GB | 7.304 | $4.75 | $4.02 | $5.48 |
A100 80GB | 8.462 | $5.50 | $4.65 | $6.35 |
H100 | 17.688 | $11.50 | $9.73 | $13.27 |
Compare features
On-Demand Training | Training Reservation | Training Hero Reservation | ||
---|---|---|---|---|
Train multi-billion-parameter models | ||||
Full model ownership | ||||
Orchestrate across multiple clouds | ||||
Scale to hundreds of GPUs | ||||
High availability | ||||
Large-scale training runs for reservation durations of 7+ days | ||||
Available Hero Run enablement services and support |
FAQ
No, Mosaic AI Model Training - pre-training is provided through the MosaicML Platform Service, not through Databricks on AWS.