Heroku Review 2026: Dyno Pricing, Eco Hours & Postgres Plans

The dyno sizes article publishes the Common Runtime table. Eco is 0.5 GB of RAM with a 1x CPU share for a $5 flat subscription. Basic is 0.5 GB for $7 a month. Standard-1X is 0.5 GB for $25 and Standard-2X is 1 GB for $50, both shared compute. Performance dynos are dedicated: Performance-M 2.5 GB for $250, Performance-L 14 GB for $500, Performance-L-RAM 30 GB for $500, Performance-XL 62 GB for $750 and Performance-2XL 126 GB for $1,500. Enterprise customers see dyno units beside each size, from 0.28 for Basic to 60 for 2XL.

The usage and billing article turns those into hourly rates: roughly $0.005 for Eco, $0.01 for Basic, $0.03 for Standard-1X, $0.06 for Standard-2X, $0.34 for Performance-M, $0.69 for L and L-RAM, $1.04 for XL and $2.08 for 2XL. Heroku computes usage from wall-clock time rather than CPU time, prorates to the second, and treats the monthly figure as the maximum for a dyno running all month. Its example is a Basic dyno scaled up at midnight and down two hours later costing about $0.02. Bills arrive for the previous month, the usage view on the Billing tab updates nightly as of 00:00 UTC, and every app counts, including staging, tutorial and test apps.

Heroku limits access to Performance dynos to customers with an established payment history, which is stated in both the sizes and tiers articles without a numeric threshold. Because Standard dynos top out at 1 GB of RAM, a new account that needs more memory for a single process cannot simply buy a Performance-M on day one, and must either run on Standard-2X or build history first. That is an unusual constraint among the platforms in our hosting and DevOps guide and worth knowing before a migration.

The Eco dyno hours article is precise. The plan costs $5 a month for a pool of 1,000 dyno hours shared by all Eco dynos in the account, renews on the first of each month, and is charged in full for the month even if you subscribe on the thirtieth; you can unsubscribe at any time and keep the remaining hours until month end. Only personal apps can use Eco; apps in a Team or Enterprise Team cannot. The app owner must be subscribed for anyone to scale an Eco dyno, and collaborators' usage counts against the owner's pool.

Hours are consumed when an Eco web dyno is awake, an Eco worker is running, or a one-off Eco dyno runs via the CLI or Scheduler, and review apps and Heroku CI count too. A web dyno with no traffic for 30 minutes sleeps, taking any Eco worker with it, and consumes nothing while asleep; it wakes after a short delay when traffic arrives. Apps with only a worker never sleep and can run all month, and monitoring pingers such as Pingdom or New Relic can keep a web dyno awake. Heroku emails at 80% and 100% of the pool; at 100% every Eco dyno in the account sleeps for the rest of the month, additional hours cannot be bought, and the documented remedy is upgrading critical apps to Basic.

The arithmetic matters. A single always-awake Eco web dyno would need 24 × 30 = 720 hours a month, leaving 280 for everything else; two always-awake apps need 1,440 and will exhaust the pool around day 20. Eco is therefore designed for apps that sleep most of the day, and the $7 Basic dyno is the right answer for anything that must stay up. Our SaaS spending guide makes the general point that a flat fee with a hard cap is cheap only when usage sits well inside the cap.

The dyno tiers article is the feature map. Metrics and language runtime metrics are available from Basic upward, not on Eco. Zero-downtime deploys arrive at Standard via Preboot and in Private, Shield and Fir via rolling deploys. Horizontal scalability, running more than one dyno per process type, begins at Standard; Eco and Basic apps can have only one dyno per process type. Autoscaling and dedicated compute begin at Performance. Dedicated networking is a Private Spaces, Shield and Fir feature. HIPAA compliance is Shield only, with Fir marked as coming. Dyno sleeping applies to Eco alone.

Read as a buying guide, the tiers say that Eco is for prototypes, Basic for a single small always-on process, Standard for anything that needs more than one web dyno or a clean deploy, and Performance for autoscaling or more than 1 GB of RAM. Heroku's own guidance in the article recommends Standard for robust apps and notes that Standard and Performance dynos can be mixed within one app. The jump from Basic to Standard-1X is $7 to $25 for the same 0.5 GB of RAM, so a team is paying $18 a month for Preboot and the ability to scale, not for more memory.

Private Spaces and Shield Spaces are separate products with their own space pricing on the Elements Marketplace and dedicated dynos priced at $125 to $1,500 and $150 to $1,800. The Teams article states that creating either requires a Team verified with at least six paid invoices, three successful charges and a minimum spend of $30, which in practice means roughly six months of paid use before an isolated network or HIPAA environment is available. Fir dynos are available only in Fir Private Spaces and inherit the same requirement.

The Postgres plans article defines tiers by tolerable downtime: Essential for apps that can tolerate up to 4 hours a month with 99.5% expected uptime, Standard for up to 1 hour, Premium for up to 15 minutes. The Elements Marketplace listing supplies prices. Essential-0 is $5 for 1 GB and 20 connections, Essential-1 $9 for 10 GB and 20 connections, Essential-2 $20 for 32 GB and 40 connections, all limited to 4,000 tables and excluding fork, follow and rollback. Standard-0 is $50 for 64 GB and 200 connections, Standard-2 $200 for 256 GB and 500 connections, Standard-3 $400 for 512 GB, Standard-4 $750 for 768 GB, Standard-5 $1,400 for 1 TB, rising to Standard-10 at $12,000 for 8 TB.

Standard adds fork and follow for replicas, continuous protection and rollback to any point in the last 4 days. Premium adds high availability with automatic failover and 7-day rollback: Premium-0 $200 for 64 GB, Premium-2 $350 for 256 GB, Premium-3 $750, Premium-4 $1,200, Premium-5 $2,500, up to Premium-10 at $24,000, with L and XL variants at some sizes offering more storage for the same RAM. Private Spaces Postgres starts at Private-0 for $300. Daily logical backups via PGBackups are free on Essential and Standard. Heroku has also introduced a Postgres Advanced tier with leader and follower pools, 100 GB of storage included per cluster and $0.20 per additional GB, on which PGBackups is not yet supported.

For a buyer the key thresholds are $5 for a database that cannot be rolled back and may be down four hours a month, $50 for one that can be restored to a point in time and replicated, and $200 for one that fails over automatically. Essential's 20-connection limit also matters: a Standard-2X web dyno running several worker processes can exhaust it, which is a reason the Standard-0 plan's 200 connections pair with Standard dynos. Our Node.js hosting guide frames the same database-tier decision across platforms.

A side project with a web process that gets a few visits a day and a nightly Scheduler job fits Eco. The bill is $5 for dynos plus $5 for Essential-0 Postgres, $10 a month or $120 a year, with the web dyno asleep most of the time and the nightly job drawing minutes from the pool. If the owner adds a Basic dyno to keep the site awake instead, the bill is $7 + $5 = $12 a month, or $144 a year, and the Eco subscription can be dropped if no other personal app needs it.

A small production app needs two Standard-1X web dynos for redundancy, one Standard-1X worker and a Standard-0 Postgres with rollback. Dynos cost 3 × $25 = $75 and the database $50, for $125 a month or $1,500 a year. The team gets Preboot deploys, Metrics, 200 database connections and 4-day rollback; it does not get autoscaling, and if a single process needs more than 1 GB it must move to Standard-2X at $50 or, once payment history allows, Performance-M at $250. Add a $9 Essential-1 database for a staging app and a Basic staging dyno at $7 and the monthly total is $125 + $16 = $141.

A scaling SaaS on two Performance-M web dynos with autoscaling, two Standard-2X workers and a Premium-0 Postgres with high availability pays 2 × $250 + 2 × $50 = $600 for dynos and $200 for the database, $800 a month or $9,600 a year, before add-ons such as the Key-Value Store, log drains or monitoring from the Elements Marketplace. Running the same formation only during a 12-hour business day would roughly halve the dyno line because of per-second proration, which is a lever the platforms with flat instance pricing do not offer. These are illustrative applications of the published prices, not quotes, and Heroku's tax treatment is not documented in the articles reviewed.

Heroku bills in arrears: each month's invoice covers the previous month's usage, computed from wall-clock time and prorated to the second, with the Eco subscription the only flat fee. Current usage is shown on the Billing tab and refreshed nightly, and Enterprise customers get a usage dashboard with CSV export. There is no published refund policy in the articles reviewed; because usage is prorated, scaling a dyno to zero stops its charges immediately, and unsubscribing from Eco leaves the remaining hours usable until month end with a full $5 charged for that month.

Teams are the collaboration layer below Enterprise. Any user can create up to five Teams, each with up to 25 members, by entering a credit card that pays for all team-owned resources at normal prorated rates; Eco dynos are unavailable to Teams, so a Team's cheapest dyno is Basic at $7. Teams get app permissions, delegated administration, unified billing, SSO through SAML identity providers and the ability to block third-party OAuth access, which are controls many platforms reserve for enterprise contracts. Hitting the five-team or 25-member limits is a sales conversation.

The verification gates are the part of Heroku's model a newcomer most needs to know. Performance dynos require an established payment history; Private and Shield Spaces require a Team with six paid invoices, three successful charges and $30 of spend. A startup that expects to need dedicated compute, autoscaling, an isolated network or HIPAA within its first few months should either start paying for Standard dynos early to build the record or plan around Standard-2X in the meantime. The SaaS evaluation checklist treats eligibility rules like these as part of the price, because they decide when a feature is actually purchasable.

Our Render review lists an Individual free plan and Team and Organization plans at $19 and $29 a month plus per-instance prices, and our Node.js hosting guide notes Render's free services spin down when idle and calls it a genuine Heroku successor with predictable invoices. Railway is usage-based, and our startup hosting guide describes it as having captured the developer experience Heroku pioneered. Fly.io runs apps in 30-plus regions on its Machines model, and DigitalOcean offers both raw droplets and a Heroku-like App Platform.

Heroku's documented position against them is clarity and maturity rather than price. It has no free tier where Render has one; its $25 entry for a scalable dyno is higher than the hobby tiers of most rivals; and it alone gates upper tiers on payment history. In return it publishes every price and limit in dated articles, bills to the second so that off-hours scaling saves money, provisions Postgres with one command across a tier ladder from $5 to $24,000, and ships Team-level SSO and permissions without an enterprise contract. A team that values a platform whose rules it can read in full will find Heroku unusually honest; a team optimising for the cheapest always-on process will find Basic competitive and everything above it expensive.

Choose Heroku if you want a mature git-push platform with buildpacks for your language, a managed Postgres you can provision and roll back without a DBA, published prices you can budget from, Team-level access controls, and the option of Private or Shield Spaces once your account qualifies. Choose Standard dynos for anything customer-facing, Basic for small always-on tools and Eco only for personal prototypes that may sleep. Skip Heroku if you need a free tier, need autoscaling or more than 1 GB per process from day one without payment history, want the lowest possible always-on price, or need HIPAA or an isolated network before you have six months of invoices.

Test it with a plan. Subscribe to Eco on the first of a month to avoid paying for days you cannot use, deploy a real app with an Essential-0 database, and watch the Billing tab's dyno-hour meter to learn how much your app actually stays awake. Try a Scheduler job and a review app to see them draw from the pool. Then move the app to a Basic dyno for a week to compare always-on behaviour, and if production is the goal, scale to two Standard-1X dynos and a Standard-0 database to test Preboot and rollback. Record the first paid invoice date, because it starts the clock on the verification Heroku requires for Performance dynos and Private Spaces.