Best overall
Highest score of the 10 providers in this ranking (100/100). A database that lives next to the app deploy.
Visit RailwayManaged Postgres, MySQL, Mongo and serverless databases. Free tiers are flagged. A $0 plan is a place to start, not a production plan.
Updated Nov 2026 · 10 providers · 2 with a published monthly plan · 9 with a free tier
Three starting points from the ranking below, each for a different kind of buyer.
Best overall
Highest score of the 10 providers in this ranking (100/100). A database that lives next to the app deploy.
Visit RailwayLowest published price
The cheapest monthly plan published in this category: Free tier, Pro $25/mo. An app that wants Postgres plus auth in one project.
Visit SupabaseBest free tier
The highest-ranked provider with a free plan or trial, after the picks above. A classic managed SQL instance inside AWS.
Visit Amazon RDSSorted by score out of 100. A from-price is the cheapest plan the vendor publishes, not a quote for your workload.
A database that lives next to the app deploy
A classic managed SQL instance inside AWS
Distributed SQL and a Postgres-compatible driver
Managed MongoDB starting on a free shared cluster
Serverless Postgres and database branches per preview
MySQL with a branch-per-deploy workflow
Edge SQLite and small replicated databases
Serverless Redis or Kafka next to an edge app
Managed Postgres or Kafka on a cloud region you choose
An app that wants Postgres plus auth in one project
You pay for compute, storage and, on serverless plans, the work each query does. These are the lines that decide the invoice.
Provisioned databases bill for an instance size per hour. Serverless ones bill for compute used and can scale to zero.
Billed per GB per month, sometimes with separate charges for I/O operations.
Point-in-time recovery, retained backups and database branches each use storage that may be billed.
A standby or read replica roughly doubles the compute line. It is often what production actually needs.
Traffic between your app and the database is cheap or free in the same region and billed across regions or clouds.
| Provider | Starting price | Billing model | Free tier |
|---|---|---|---|
| Railway | Hobby $5/mo plus usage | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| Amazon RDS | Instance pricing, limited free tier | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| CockroachDB | Free serverless tier, then usage | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| MongoDB Atlas | Free M0, then paid tiers | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| Neon | Free tier, then usage | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| PlanetScale | Free hobby, paid plans | Fixed plans with a published monthly price. Hourly billing may apply up to that cap. | Yes |
| Turso | Free tier, then paid plans | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| Upstash | Free tier, then per request | Pay for what you use: requests, tokens, traffic or compute time. The bill moves with volume. | Yes |
| Aiven | Trial, then monthly plans | Fixed plans with a published monthly price. Hourly billing may apply up to that cap. | No |
| Supabase | Free tier, Pro $25/mo | Fixed plans with a published monthly price. Hourly billing may apply up to that cap. | Yes |
You want to ship a Postgres app without managing a server, and you accept scale-to-zero or usage billing. Neon and Supabase are typical starting points.
You want a database that stays up like a VM-backed service, with familiar sizing. Amazon RDS and Aiven fit teams that already think in instance classes.
Schema change workflow matters as much as hosting. PlanetScale’s branching model is the reason to shortlist it over a generic MySQL host.
You need SQLite at the edge (Turso), serverless Redis (Upstash) or a document store (MongoDB Atlas) beside the primary app database—not as a drop-in Postgres replacement.
The useful comparison is the first paid plan plus what the free tier actually allows. Neon, Supabase, Turso and MongoDB Atlas all let you start without designing a server. Amazon RDS and Aiven are closer to a classic managed instance. PlanetScale is MySQL with a branching workflow. CockroachDB, Upstash and Railway cover distributed SQL, serverless Redis and app-adjacent databases. They do not replace each other.
Watch the limits that wake you up: compute hours, storage, row reads, connection counts and egress. A free Postgres that sleeps is a fine prototype and a bad payments database. Neon and similar serverless Postgres products make scale-to-zero explicit—read the wake behavior before you put a synchronous checkout path on it.
Backups, point-in-time recovery and a region you can name should be on the pricing page of any plan you call production. Amazon RDS and Aiven feel familiar if you already think in instance sizes. Supabase bundles auth and APIs with Postgres; that is convenient, but the database limits still need their own checklist.
Engine choice is sticky. Postgres, MySQL, SQLite-at-the-edge, Redis and document stores imply different query habits and migration costs. Turso fits edge SQLite patterns; Upstash fits serverless Redis; MongoDB Atlas fits document workloads. Switching later is a migration project, not a toggle in the panel.
Exit paths matter. Standard dumps, logical replication or a documented export beat a platform you cannot leave. Test a restore on the first paid plan, not only a happy-path insert on the free tier. If the app lives on a VPS or cloud hosting account, put the database in a region that matches the app’s latency budget.
This ranking highlights public free tiers and the first paid plan you can read without a sales call. It is not a TPC benchmark and it does not claim query latency winners. Pilot with your schema and your connection pool settings.
Step 1
Postgres, MySQL, SQLite at the edge, Redis, or a document store. Switching later is a migration, not a toggle.
Step 2
Storage, compute hours, pause-on-idle and connection caps. Write them down before you celebrate $0.
Step 3
The plan with backups, a region and enough connections. That number is the comparison, not the hobby tier.
Step 4
Standard dumps and a documented export beat a platform you cannot leave.
Every query pays the latency, and the transfer can be billed.
Serverless is great for spiky or idle workloads. Constant load can cost more than a fixed instance.
A backup you have never restored is a guess. Test a restore before launch.
Check that you can export your data in a standard format and how long it takes.
Each provider is scored out of 100 against the other 9providers on this page, using only what is published on the vendor's site. The sponsor slot is not an input.
Full editorial policyFor a Postgres app, start with a serverless Postgres if you want to ship today, or a classic managed instance if you want a server that does not sleep. The ranking shows public commercial terms, not a query benchmark.
Enough to build and to demo. Production needs backups, a region and a plan that stays awake. Read the limits on the vendor page.
No. Sponsored slots are labeled and are not part of the score.
Pick the engine your app and team already know. Migration between engines is a project. Compare vendors inside one engine, not across engines on price alone.
Fine for prototypes and low-traffic apps. Risky for synchronous user paths that cannot wait for a cold start. Read wake behavior before production.
You can self-host on a VPS, but then backups and failover are yours. Managed plans exist so those jobs have a vendor SLA and a restore path.