Amazon Nova Pro: Features, Pricing, and Deployment
Nova Pro is an active Amazon Bedrock model for applications that combine text with images or video. Its 300,000-token context window suits document analysis, media review, and codebase questions where the input is larger than a typical chat. Evaluate it by region, latency, token cost, and results on your own repository.
What Is Amazon Nova Pro?
The Bedrock model ID is amazon.nova-pro-v1:0. It belongs to the first Amazon Nova generation and is positioned as a balanced multimodal model. It is active in Bedrock, but it is not Amazon's current flagship. Amazon maintains separate documentation for Nova 2.
Access runs through Amazon Bedrock rather than a downloadable weight package. It supports the Invoke and Converse APIs.
Multimodal Inputs and Text Output
The model accepts text, images, and video, then produces text. It can review a screenshot with a bug report or summarize a recorded product flow. It does not generate image, video, or audio output.
The 300K limit is a capacity boundary, not a reason to send every file. Start with the material needed for the decision and measure whether more context improves the answer.
Nova Pro vs Claude on Bedrock: What to Compare
"Claude on Bedrock" describes a changing model family, so select an exact Claude model ID before testing. Keep the region, repository snapshot, tools, and pass conditions fixed.
| Decision point | Amazon model route | Claude model route |
|---|---|---|
| Input mix | Text, image, and video input | Check the selected Claude model card |
| Context and reasoning | 300K context; use documented controls | Varies by the selected Claude version |
| Bedrock access | Invoke and Converse APIs | Confirm APIs for the chosen model |
| Cost | Check the live Bedrock price table | Check the same region and service tier |
| Evaluation | Test repository outcomes and latency | Run the identical task and scoring method |
This avoids comparing a fixed model with an unspecified family.
Deployment Options in Amazon Bedrock
Amazon Bedrock offers Standard, Priority, and Flex service tiers for Nova Pro. Region and inference choices affect request location and available throughput. Confirm the target region before building an endpoint.
Standard on-demand pricing starts at $0.80 per 1M input tokens and $3.20 per 1M output tokens. Rates vary by region and service tier. The maximum output is 5K tokens.
Record the model ID, region, service tier, timeout, and retry policy in configuration. This makes tests reproducible and keeps routing changes out of application logic.
Coding Evaluation and Benchmark Limits
Published code-generation benchmarks cover controlled tasks. They do not show how the model will navigate your repository, call tools, respect local conventions, or recover from a failing test. A practical coding evaluation should include:
- Ask it to repair a known bug and require the failing test before the patch and the passing test afterward.
- Give it a multi-file refactor with a fixed public interface and inspect every changed dependency.
- Test a screenshot-backed UI issue to see whether the image input improves the diagnosis.
- Repeat the task at two context sizes and compare success, latency, token use, and reviewer time.
Keep the harness unchanged when testing another model.
Nova Pro vs Nova Lite v1
The two version 1 models accept text, images, and video, return text, and support 300K context. Pro is the higher-capability choice. Lite is designed for lower cost and latency.
Test Lite on high-volume work first. Move a task to Pro when the measured quality gain justifies the cost or response time. Use the current regional Bedrock price table for the final estimate.
Checking Amazon Nova Pro Access in Verdent
Amazon Nova Pro is not included in Verdent's built-in model lineup, and Verdent BYOK does not offer Amazon Bedrock as a provider.
Check the model selector and current documentation before planning a Verdent workflow around it. If the ID is absent, treat the Bedrock deployment and the Verdent workflow as separate routes.
Validate the Model Route Before You Commit
Confirm the exact ID in Verdent's current model catalog. If you plan to bring a provider key, review the documented BYOK providers before you build the evaluation around Verdent.
Frequently Asked Questions
What context and input types does Amazon Nova Pro support?
It supports a 300,000-token context window with text, image, and video input, and it returns text. Practical prompt size should be set by task quality, cost, and review time rather than the maximum alone.
Is Amazon Nova Pro Amazon's current flagship model?
No. It is an active, balanced model from the first Nova generation. Amazon also maintains a separate Nova 2 family, so model selection should use the exact ID and current family documentation.
How should I choose between Nova Pro and Nova Lite v1?
Both accept the same main input types and support 300K context. Test Lite for cost-sensitive or latency-sensitive work, then use Pro when a measured capability gain is worth the difference.
Is Amazon Nova Pro built into Verdent?
Amazon Nova Pro is not built into Verdent, and Amazon Bedrock is not a Verdent BYOK provider. Check the live selector before designing a workflow around this model.