Upstash vs Redis Cloud: Free Tiers, Pricing & Limits

A rate limiter, response cache and session store reach different limits. A rate limiter may use several Redis commands for one application request; a cache may exhaust transfer before storage; a session store may need more memory even when traffic is low. Monthly commands alone cannot tell you whether a plan fits.

Use the workload calculator on this page to compare the published configurations it supports. A plan marked unsuitable should not be treated as a cheap recommendation. A Redis Cloud paid configuration needs a provider quote or console selection; its starting price is not a universal capacity guarantee.

Pricing checked 9 September 2026; values are USD before tax. Upstash Free includes 256 MB data, 500,000 monthly commands and 10 GB monthly bandwidth. Redis Cloud Free provides 30 MB with separate throughput, connection and transfer limits. Free and Essentials are shared deployments; Redis Cloud Pro is dedicated.

Upstash Pay as You Go charges $0.20 per 100,000 commands. The Free plan command allowance does not carry into the paid plan. Storage and transfer have their own allowances and rates; low traffic does not automatically mean no bill. Fixed plans remove per-command billing but retain capacity and throughput constraints.

Keep application requests separate from billable commands. Read-modify-write logic, retries and queue polling can increase operations. A pipeline reduces network round trips without necessarily reducing billed commands. Additional regions can change write replication and storage costs. The calculator models one region without production add-ons; verify a multi-region design separately.

These are the calculator’s preset scenarios: planning inputs, not observed customer bills. Keep peak operations and simultaneous connections consistent with your application; a monthly average can hide a burst that exceeds a plan limit.

The rate limiter preset uses 300,000 monthly commands, 10 MB data and 1 GB transfer. Its Upstash PAYG command component is $0.60; the Free plan may be preferable if every constraint fits. The application cache preset uses 10 million commands, 500 MB data and 250 GB transfer: $20 in commands plus (250 − 200) × $0.03 = $1.50 transfer, for a $21.50 PAYG estimate. This assumes billable storage remains within the first 1 GB allowance. The session store preset uses 2 million commands, 200 MB data and 20 GB transfer, giving a $4 PAYG command component. Paid commands are charged from the first billable command; the 500,000-command Free allowance does not carry over.

These examples use one region without production add-ons. Change the command volume to match your implementation: application requests can trigger several Redis operations, and cache misses, retries and session refreshes affect the total. Measure transfer and stored data in the provider dashboard, including overhead. Use the calculator result as an estimate and review production requirements before selecting a plan.

Upstash provides REST access as well as Redis-protocol connectivity. HTTP is useful for a runtime that cannot keep a TCP connection open. A persistent Node.js API can instead use an appropriate Redis client and connection pool. Redis Cloud is generally accessed through Redis-protocol clients.

Choose a database region close to the code that reads and writes it, not just close to the browser. An edge frontend may still call an API and database in one region. Verify round-trip latency with your payloads and concurrency; this page makes no universal fastest-provider claim.

Do not treat Redis compatibility as a yes/no label. Inventory the commands issued by your client and libraries, including Lua scripts, transactions, streams, JSON and search. Upstash supports JSON and its own SEARCH.* commands, but SEARCH.* is not API-compatible with RediSearch FT.*. A library expecting FT.CREATE or FT.SEARCH needs explicit verification or adaptation.

Create a small compatibility test around the operations your application actually uses. Include expiration, conditional writes, atomicity and error handling. Passing a GET/SET smoke test says little about a queue library or a search integration.

Persistence, backups, multi-zone availability and an uptime SLA are different features. Check each on the selected plan. A cache that can be rebuilt has a different recovery requirement from login sessions or a job queue. Do not compare a basic shared configuration with an added high-availability deployment as if their service guarantees matched.

Upstash PAYG supports a monthly budget cap; reaching it can rate-limit the database. Fixed plan storage or transfer limits can also trigger an upgrade or limit service, depending on settings. Decide whether limiting spend or maintaining service takes priority, and configure alerts before relying on the database.

Keep the existing provider available until the replacement passes application checks. A client URL change is only one part of moving a cache, session store or queue.

Stay with the current provider when the workload fits and changing it offers no clear operational benefit. Upgrade within the provider when the issue is a capacity limit that the next plan solves. Switch when a required connection model, command API or deployment feature is missing.

For the rest of the application stack, compare Railway vs Vercel and Railway vs Render. Include the application host’s outgoing traffic charges as well as the Redis bill when the services are on different providers.