When will DeepSeek V4 Pro be available on Amazon Bedrock?

AWS tends to prioritize models based on customer demand, so it may be worth raising this with your AWS account team and asking them to submit an internal request. That probably won't guarantee a launch date, but it puts the demand on their radar.

If keeping the charges on your AWS bill is the main requirement, look at SageMaker endpoints, AWS Marketplace offerings, or deploying the open weights yourself. Those options can provide access sooner, but they usually involve instance-hour pricing, so you pay for the endpoint even when usage is low. Compare the cost against your actual daily traffic before choosing self-hosting.

A model gateway can be a good compromise. Put a thin internal proxy or something like LiteLLM in front of Bedrock and external providers, then keep one interface for authentication, usage tracking, and observability. You won't necessarily get every provider folded into the AWS bill, but you can avoid rewriting your application whenever a model is unavailable on Bedrock.

If speed matters more than billing consolidation, using a direct DeepSeek-compatible provider is probably the simplest route. You can also check other AWS regions, since cross-region inference may sometimes expose a model before it appears in the region you normally use. Just verify data-residency, latency, and pricing requirements first.

The concern is that repeatedly submitting requests doesn't solve the release-cadence problem. If a model needs to be available close to launch for it to be useful, waiting for an internal request process may not be practical. For models that need to be available immediately, using the provider directly or a third-party inference platform is likely more realistic.