Methodology

How the options are ranked

Every option is rated on the same six criteria, from 0 to 3, using only what the vendor publishes on its own pricing pages, docs and trust pages. Each rating is weighted, and the weighted ratings add up to a score out of 100. Nothing here comes from paid placement, user reviews or our own benchmarks.

Last reviewed

The rules

  • Same criteria for everyone. Every option is judged on the six criteria below, against the same written scale.
  • Only published facts. A rating rests on what the vendor publishes. Where a fact isn't published, we say so and rate on what is, and the entry tells you where to check.
  • Best published mode. Where an option offers more than one way to run (for example its own cloud or yours), we rate the best mode available on a published plan and say which mode that is.
  • The score. Each rating counts as a share of its criterion's weight: a 3 earns the full weight, a 2 two-thirds, a 1 one-third, a 0 nothing. The weights add up to 100.
  • Ties. Equal scores are split by the better "where apps run" rating, then by compliance, then alphabetically.
  • Reviews. All prices and claims are rechecked on each review; the date is at the top of every ranking page.

The six criteria

Where apps run weight 25

Can your apps and data run inside a cloud account you own?

Why it matters. Where your containers and databases live decides who controls the network, who holds the data, whether your cloud credits and committed-spend discounts apply, and what a customer's security review will ask. It is also the hardest thing to change later, so it carries the most weight.

How it was judged. Read from each option's own docs and pricing page. We rate the best hosting mode the option offers on a published plan, and say which mode that is.

0Vendor-hosted only, with no choice of region.
1Vendor-hosted, with a choice of regions.
2In your own account on one cloud, or in your own cloud only through an enterprise contract.
3In your own account on more than one major cloud (AWS, Google Cloud, Azure), on a published plan.

Compliance weight 20

Can you get a SOC 2 report and sign a HIPAA BAA, and on what terms?

Why it matters. If you sell to healthcare, finance or larger companies, the platform's own attestations and its willingness to sign a Business Associate Agreement (BAA) decide whether you can use it at all. Paying for a compliance tier you have to negotiate is a different cost from one with a public price.

How it was judged. Read from each option's pricing page, trust or security page and compliance docs. We do not audit anyone; we record what the vendor publishes and where.

0No SOC 2 report or HIPAA terms published.
1A SOC 2 report is published, but HIPAA terms for this product are not.
2SOC 2 and HIPAA are available, but only on an enterprise or custom contract, or as an add-on without a public price.
3SOC 2 and a HIPAA BAA (or HIPAA-ready infrastructure) are available on a published plan or at a published price.

GPU support weight 15

Can you run your own GPU workloads (inference, fine-tuning) on it?

Why it matters. Many teams now ship a model next to their app. A PaaS that cannot run GPU containers pushes that part of the stack somewhere else, with its own deploys, networking and bill.

How it was judged. Read from each option's pricing page and docs. Hosted model APIs do not count: we ask whether you can deploy your own container onto a GPU.

0No GPU instances for your own workloads.
1GPUs offered in a limited or scaled-back way.
2GPUs available, but on one cloud only or with notable limits.
3GPU instances documented and self-serve for your own workloads.

Cost model weight 15

Are the prices public, and does the bill follow what you actually use?

Why it matters. A price you can read before signing up is a price you can check. Billing by usage rather than by fixed boxes or seats keeps the bill close to what your app needs, and avoids paying for headroom.

How it was judged. Read from each option's public pricing page on the review date. We rate the shape of the pricing, not whether it is cheap: cheapness depends on your workload, so we give the published rates and leave the sums to you.

0No public prices.
1Public base prices, but the tiers most teams need (private networking, compliance) are quoted by sales or carry a steep premium.
2Public prices, billed by fixed instance sizes, seats or flat plan fees.
3Public rates billed by actual usage (per second or per resource), with no seat fees.

Ease of exit weight 15

How hard is it to leave: if you do, what keeps running and what must be rebuilt?

Why it matters. Every platform gets left eventually: a price change, an acquisition, a requirement it can't meet. The cheaper the exit, the less the platform can charge you for staying.

How it was judged. Judged from how apps are packaged (standard containers or buildpacks versus platform-specific runtimes), where the infrastructure lives, and what the vendor states about leaving. This rates exit difficulty the other way round: a higher rating means an easier exit.

0Apps depend on platform-only runtimes; leaving means a rewrite.
1Apps lean on platform-specific features you would have to replace.
2Standard containers or buildpacks; you move images, config and data to a new host.
3The infrastructure is already yours: if you stop paying, it keeps running in your account.

Operations handled weight 10

How much running of infrastructure is left to you?

Why it matters. This is why anyone picks a PaaS. It carries the least weight here only because every option on the list does a lot of it; the differences are in the edges.

How it was judged. Judged from what the option sets up for you out of the box: builds from Git, deploys, TLS, autoscaling, logs, previews, and whether you also manage a cloud account.

0You assemble most of the platform yourself.
1Runs your containers; you wire up CI/CD, networking and the rest from other services.
2Most of the platform is handled, but you also own a cloud account, a CLI-first workflow or a cluster.
3Push code and the platform handles the rest; nothing else to own.

Why these weights

The weights favour decisions that are expensive to undo. Where apps run carries the most because moving data and networks between clouds is the hardest migration a team makes. Compliance comes next because it decides whether you can use a platform at all. GPU support, cost model and exit difficulty share the middle. Operations handled counts least, not because it is unimportant, but because every option here handles most of it; the differences are at the edges.

These weights suit a team that expects to grow, sell to larger customers, or run models next to its app. A solo developer shipping a side project would weigh things differently, and should.

Reweigh it for your case

Move the sliders to set what matters to you. The ratings stay exactly as published on the ranking page; only the weights change.

Your weights

Set a criterion to 0 to ignore it. The ratings stay the same; only how much each one counts changes.

Our ranking

  1. Porter 96.7
  2. Northflank 85
  3. Qovery 80
  4. Google Cloud Run 80
  5. Railway 65
  6. Fly.io 65
  7. Vercel 61.7
  8. Render 58.3
  9. Heroku 53.3
  10. DigitalOcean App Platform 45

What this ranking doesn't measure

  • Speed, uptime or support quality: these need long-running measurement we haven't published.
  • Which is cheapest for your app: that depends on your workload. Every entry gives the published rates instead.
  • Ease of use beyond what each platform sets up for you, which is a matter of taste.

Spotted an outdated fact?

Prices and plans change. If something here no longer matches the vendor's own page, tell us what changed and link the source. We check it before the next review.