AWS announced on November 9, 2026, that Amazon DocumentDB Serverless has a new platform version, numbered 4, which can deliver up to 35% more throughput than the previous version 3. New serverless clusters start on version 4 by default wherever it is offered, while existing clusters get it as a pending maintenance action that the owner decides when to apply.
For teams running MongoDB-compatible workloads on DocumentDB Serverless, this is a performance gain that needs no code change and no new instance class. The catch is that the move is one-way: once a cluster runs version 4, AWS does not let it go back to version 3, so a short test before upgrading production is worth the effort.
What AWS Changed
DocumentDB Serverless bills and scales compute in DocumentDB capacity units, or DCUs, where one unit is roughly 2 GiB of memory with matching CPU and network. A platform version is the underlying serverless layer that AWS manages for a cluster. Each new version can bring better performance, faster scaling or extra features.
According to the DocumentDB developer guide, the two versions in use today compare like this:
- Platform version 3: capacity from 0.5 to 256 DCUs, treated as the baseline for performance.
- Platform version 4: the same 0.5 to 256 DCU range, with up to 35% higher throughput than version 3.
- Version 3 remains available in every region that offers DocumentDB Serverless.
- Version 4 is listed for most commercial regions, including N. Virginia, Ohio, Oregon, Mumbai, Hyderabad, Frankfurt, Ireland and London, plus both China regions and both GovCloud regions.
The capacity range is unchanged, so the gain comes from doing more work per DCU rather than from bigger limits. AWS phrases it as an upper bound, which means real results will depend on the shape of each workload.
How the Upgrade Works
Clusters created after the release automatically use the newest platform version their region offers. Existing clusters on version 3 show a pending maintenance action called serverless-platform-version-update, and AWS leaves the timing to the customer.
The documentation describes a short sequence for checking and applying it:
- Open the cluster in the DocumentDB console and read the Platform version field under Configuration.
- Or read the ServerlessV2PlatformVersion field returned by the describe-db-clusters command or the DescribeDBClusters API.
- Run describe-serverless-v2-platform-versions with the default-only flag to see which version new clusters receive in a region.
- Call describe-pending-maintenance-actions with the cluster ARN to confirm the upgrade is waiting.
- Apply it with apply-pending-maintenance-action, choosing an opt-in type of immediate or next-maintenance.
Choosing next-maintenance lets the change happen inside the cluster's normal maintenance window. Choosing immediate starts it right away, which suits test and staging environments where timing is less sensitive.
What It Means for Developers
Higher throughput per DCU can matter in two ways. Under the same load, a cluster may need fewer capacity units, which can lower the DCU-hour bill. Under peak load, the same maximum capacity setting may now absorb more traffic before it becomes a bottleneck.
That makes this a good moment to revisit scaling settings. If your writer often sits at its MaxCapacity with DCUUtilization near 100%, measure again after the upgrade before raising the limit. If the cluster rarely gets close to its ceiling, you may be able to trim the maximum and keep a tighter budget cap.
Remember that some instance limits, such as active connections, open cursors and open transactions, follow the current capacity of a serverless instance. Lowering capacity settings too far after the upgrade can bring those limits into play even if raw throughput looks fine.
If you also build AI features on MongoDB-style data, our story on the MongoDB Vercel AI SDK integration covers another option in the same space.
How to Get Started
Start with a non-production cluster that mirrors your real traffic. Record baseline numbers from CloudWatch, mainly serverlessDatabaseCapacity, DCUUtilization and query latency, then apply the upgrade and run the same load again.
Because there is no downgrade path, take a fresh snapshot and agree on the maintenance window before you touch production. Use the next-maintenance opt-in for busy systems so the change lands at a planned time.
Infrastructure-as-code users should also check their drift detection. Scripts that compare cluster settings may start to report the new platform version field, and new environments will pick up version 4 by default in supported regions.
Conclusion
Platform version 4 gives Amazon DocumentDB Serverless up to 35% more throughput within the same 0.5 to 256 DCU range, with new clusters getting it automatically. Existing users control the timing through a standard maintenance action, but should test first, since the upgrade cannot be reversed.
Sources
- AWS What's New: Amazon DocumentDB Serverless now provides up to 35% higher throughput
- Amazon DocumentDB Developer Guide: How DocumentDB serverless works
- Amazon DocumentDB Developer Guide: Managing DocumentDB serverless
See my articles more often when you search Google.
