Generative AI platforms & models · AWS vs Google Cloud
Amazon Bedrock model customization vs Gemini model tuning
Bedrock model customization maps to Gemini tuning on Agent Platform: supervised fine-tuning and preference tuning for Gemini, plus tuning of open models from Model Garden.
Side-by-side comparison
| Aspect | Amazon Web Services | Google Cloud |
|---|---|---|
| Service | Amazon Bedrock model customization | Gemini model tuning |
| Lifecycle | Generally available | Generally available |
| What it is | Supervised fine-tuning, reinforcement fine-tuning (reward functions in Lambda), and distillation for supported Bedrock models. | Supervised fine-tuning and preference tuning for Gemini models, embedding tuning, and tuning of open models from Model Garden. |
| Pricing model | Billed per training token (tokens in corpus × epochs) plus monthly model storage. Inference on a custom model may require Provisioned Throughput or on-demand custom model pricing, depending on the model. | Tuning jobs are billed for the training work (see the tuning section of the pricing page). Calls to a tuned model are billed per token. |
| Docs | Amazon Bedrock model customization docs | Gemini model tuning docs |
| Pricing page | Amazon Bedrock model customization pricing | Gemini model tuning pricing |
Key differences
- Tuning targets Gemini and selected open models; which methods a model supports varies by model.
- Google recommends trying prompt design and grounding first and documents when each tuning method fits.
- Tuned Gemini models are served by the platform without you provisioning endpoints for them.
Pricing summary
Pricing models only. No prices are quoted because they change by region and over time.
- AWS
- Billed per training token (tokens in corpus × epochs) plus monthly model storage. Inference on a custom model may require Provisioned Throughput or on-demand custom model pricing, depending on the model. Amazon Bedrock model customization pricing →
- GCP
- Tuning jobs are billed for the training work (see the tuning section of the pricing page). Calls to a tuned model are billed per token. Gemini model tuning pricing →
And on Azure?
Direct match Azure's equivalent is Foundry fine-tuning. AWS vs Azure comparison →
FAQ
What is the Google Cloud equivalent of Amazon Bedrock model customization?
Bedrock model customization maps to Gemini tuning on Agent Platform: supervised fine-tuning and preference tuning for Gemini, plus tuning of open models from Model Garden.
How does pricing compare between Amazon Bedrock model customization and Gemini model tuning?
AWS: Billed per training token (tokens in corpus × epochs) plus monthly model storage. Inference on a custom model may require Provisioned Throughput or on-demand custom model pricing, depending on the model. Google Cloud: Tuning jobs are billed for the training work (see the tuning section of the pricing page). Calls to a tuned model are billed per token. Check the official pricing pages for current rates.
What is the Azure equivalent?
Foundry fine-tuning (direct match). Bedrock model customization maps to Foundry fine-tuning.