Upstash pricing

About

Upstash is a serverless data platform offering managed Redis, a Vector database, the QStash message queue / scheduler, a durable Workflow engine, and a full-text Search product — all billed on a request-metered, pay-only-for-what-you-use model. Founded in 2020 and headquartered in California, Upstash targets developers building on serverless and edge runtimes (Vercel, AWS Lambda, Cloudflare Workers, Fly) who want a database that scales to zero and charges per operation rather than per provisioned instance.

The company positions itself against both incumbent managed-Redis vendors (Redis Cloud, AWS ElastiCache) and serverless-native data peers, competing on the granularity of its billing: a free tier on every product, a per-request “Pay as You Go” plan for variable traffic, and flat-rate “Fixed” plans for teams that want predictable monthly costs. Each product is priced and packaged independently, so a customer can run Redis on a Fixed plan while paying QStash per message.

Upstash’s distinguishing commercial idea is that serverless data should bill like serverless compute — by the operation, with no idle cost — and that production-grade guarantees (SLA, multi-zone HA, SOC-2, encryption at rest) are sold as an add-on (“Prod Pack”) layered on any paid plan rather than locked behind an enterprise tier.

Pricing summary : per-request serverless rates with optional flat-rate Fixed plans

Upstash uses per-request usage-based pricing as the default across all five products: every product has a $0 free tier, then a “Pay as You Go” plan that meters the core operation (commands, requests, messages, steps, or queries). For teams with steady load, each product also offers Fixed plans — a flat monthly price that removes per-request billing in exchange for capped capacity. Enterprise is custom-quoted.

The billing dimensions in play:

• Core operation metering — Redis $0.2 / 100K commands, Vector $0.4 / 100K requests, QStash $1 / 100K messages, Workflow $1 / 100K steps, Search $0.05 / 1K requests.
• Storage — $0.25 per GB on Redis (first 1 GB free), Vector, and Search Fixed/Pro plans (daily-average basis).
• Bandwidth — free up to a threshold (200 GB on Redis/Vector; 50 GB on QStash/Workflow), then $0.03/GB (Redis/Vector) or $0.05/GB (QStash/Workflow).
• Fixed monthly plans — Redis $10–$1,500/mo by data size, Vector $60/mo, QStash/Workflow $180 (1M) and $420 (10M).
• Per-region multiplier — Redis Fixed plans add a per-read-region surcharge ($5–$750/region matching the tier); global write commands are re-counted per replica region.
• Prod Pack add-on — +$200/mo per database adds SLA, multi-zone HA, SOC-2, encryption at rest, Prometheus/Datadog.

What makes this different: production guarantees are an unbundled $200/mo add-on rather than an enterprise-tier gate, and the same product can be switched between per-request and flat-rate billing without changing vendors.

Pricing by product

Each Upstash product is priced independently. The hero grid above summarizes the cross-product model; the tables below give the per-product detail. Prices are USD.

Redis (per-request and Fixed plans)

Prod Pack add-on: +$200/mo per database (uptime SLA, multi-zone HA, encryption at rest, SOC-2, Prometheus, Datadog). Available on any paid plan.

Vector (per-request, Fixed, and Pro)

QStash (message queue / scheduler)

Prod Pack add-on: +$200/mo (uptime SLA, encryption at rest, SOC-2, Prometheus, Datadog).

Workflow (durable execution, built on QStash)

Prod Pack add-on: +$200/mo (uptime SLA, encryption at rest, SOC-2, Prometheus, Datadog).

Search (full-text + semantic, Preview)

Sales motions across products: PLG / self-serve for Free, Pay as You Go, and Fixed plans (credit-card upgrade in dashboard); sales-led for Vector Pro and all Enterprise tiers (sales@upstash.com).

Hidden costs : storage, bandwidth, retries, and per-region replication

Upstash’s headline per-request rates are only one line of the bill. The costs that surprise teams are the ones that sit outside the per-command meter — storage, bandwidth, retries, and especially the per-region multiplier on global Redis. Two representative archetypes show how the real bill builds up. Prices are USD.

A session-cache startup running a global Redis database (1 primary + 1 read region) at moderate volume on Pay as You Go:

The lesson: the per-region replication surcharge and the $200 Prod Pack add-on dwarf the raw command cost — the $100 of commands becomes a ~$351 production bill once you add a read region and turn on the SLA. A team that assumed “$0.2 per 100K” was the whole story under-budgets by ~3.5×.

A background-jobs SaaS leaning on QStash for webhooks with a flaky downstream endpoint, on Pay as You Go:

The lesson: every retry is a billed message, so an unreliable endpoint inflates the QStash bill silently — the 2M retries add 25% to the base cost, and active schedules beyond the first 1,000 are billed separately at $0.01 each. Compare this packaging discipline with the resolution-metered model on Intercom’s Fin, where the billable event is an outcome rather than an attempt.

Want to estimate your own Upstash bill? Use the Upstash pricing calculator to model your monthly cost based on commands, requests, storage, bandwidth, and per-region replication.

Pricing evolution : from serverless Redis to a five-product per-request platform

Upstash’s pricing has evolved by adding products and lowering entry prices, not by raising per-request rates. The core Redis rate ($0.2 per 100K commands) has held since at least 2023; the meaningful changes are new products (QStash, Vector, Workflow, Search), the deprecation of Kafka, and a 2025 repackaging that slashed the flat-rate entry price.

Cadence

Tracked range: 2021–2026. Quarters not listed were verified stable (0 price changes, 0 SKU additions) against Wayback snapshots of upstash.com/pricing and the Upstash blog.

Notable changes

• 2022-07-18 — QStash launches as a serverless message queue/scheduler at $1 per 100K messages, with each delivery attempt (including retries) billed as one message (upstash.com/blog/qstash-announcement).
• 2024-01-31 — Upstash Vector launches at $0.4 per 100K requests; the Feb 2024 Wayback snapshot shows the product tab order Redis · Kafka · Vector · QStash (upstash.com/blog/introducing-vector-database).
• 2024-09-06 — Workflow launches on top of QStash, and Upstash announces it will discontinue Serverless Kafka over the following six months (upstash.com/blog/workflow-kafka).
• 2025-03-11 — The Redis repricing: old “Pro 2K” ($280/mo) and “Pro 10K” ($680/mo) plans are retired in favor of Fixed plans from $10/mo, with a larger monthly free tier and bigger PAYG limits (upstash.com/blog/redis-new-pricing).
• 2025-10-03 — QStash and Workflow declared GA / production-ready (upstash.com/blog/qstash-qa).

The 2025 Redis repricing in detail

The March 2025 change is the single largest pricing move in Upstash’s history, and Wayback corroborates both ends of it. The September 2023 snapshot of upstash.com/pricing shows a Redis flat-rate ladder of “Pro 2K” at $280/mo (+$100 per read region) and “Pro 10K” at $680/mo (+$200 per read region), with a free tier capped at 10,000 commands per day. The current snapshot shows that ladder replaced by Fixed plans from $10/mo to $1,500/mo and a free tier of 500K commands per month.

That is roughly a 28× reduction in the entry flat-rate price ($280 → $10). Under the contract’s >5× rule the move requires a secondary source, which the official changelog post supplies directly: Upstash’s own “New Pricing and Increased Limits for Upstash Redis” post (2025-03-11) confirms the $10 starting Fixed plan, the move from 10K commands/day to 500K commands/month on the free tier, and the lift of Pay-as-You-Go storage from 10GB to 100GB. The per-request rate ($0.2 / 100K commands) did not change — Upstash lowered the floor to compete for small production workloads while leaving its usage meter intact.

What’s unique : scale-to-zero data priced like serverless compute

Per-operation billing with scale-to-zero. Upstash’s founding idea is that a database should bill like a serverless function — by the operation, with no idle cost. A Redis database that receives no traffic costs $0; one that receives 100K commands costs $0.20. This is the pure-usage pricing model applied to stateful data, and it is what differentiates Upstash from provisioned managed-Redis vendors that charge for an always-on instance.

Production guarantees as an unbundled add-on. Rather than gating an SLA, multi-zone HA, SOC-2, and encryption at rest behind a contact-sales enterprise tier, Upstash sells them as a flat $200/mo Prod Pack that snaps onto any paid plan. A two-person startup on Pay as You Go can buy the same production guarantees a large enterprise gets, without a sales call — a notably democratic packaging choice for infrastructure pricing.

A budget cap that actually caps the bill. On Pay as You Go you set a maximum monthly spend per database; beyond it the database is rate-limited rather than overage-billed, so the bill is mathematically incapable of exceeding the cap. This directly addresses the bill-shock problem that plagues pure-usage pricing and is rare among usage-metered infrastructure vendors.

Per-product, switchable billing modes. Each of the five products is priced independently, and within each product a customer can flip between per-request (Pay as You Go) and flat-rate (Fixed) billing month to month without migrating vendors. A team can run Redis on a predictable Fixed plan while paying QStash purely per message.

Honest meter design — retries and replicas are counted. Upstash bills every delivery attempt (including retries) on QStash/Workflow, and re-counts global Redis writes per replica region. The mechanics are unusually well-documented in the pricing FAQ, but they are real cost amplifiers that buyers must model — see Hidden costs.

Strengths & weaknesses

Billing UX : budget caps, soft limits, and dashboard self-upgrade

Upstash exposes several named billing controls in its dashboard and docs:

• Budget cap (Pay as You Go) — set a maximum monthly spend per database; usage beyond the cap is rate-limited so the bill never exceeds the chosen budget. Email alerts fire at 70% and 90% of budget.
• Auto-upgrade (Fixed plans) — when a Fixed-plan database hits its bandwidth or storage limit, it can auto-upgrade to the next tier; if disabled, the database is rate-limited (writes blocked on storage cases, traffic blocked on bandwidth cases) instead of overage-billed.
• Soft limits with contact-on-overage — QStash/Workflow “messages/steps per day” and bandwidth are soft limits; sustained overage triggers an upgrade conversation (Fixed 1M → Fixed 10M → Enterprise) rather than an automatic charge, with 429 errors only as a last resort.
• Prod Pack toggle — production guarantees (SLA, multi-zone HA, SOC-2, encryption at rest, Prometheus/Datadog) are enabled per database from the dashboard details page for +$200/mo, on any paid plan except Free.
• Per-region cost surfacing — Redis Fixed plans show the read-region surcharge ($5–$750/region) inline, and the docs flag that global write commands are re-counted per replica region (e.g. 1 primary + 1 read region = $0.4 per 100K writes).
• Credit-card self-upgrade — entering a card in the console upgrades a free database to Pay as You Go automatically and lifts the free-tier limits.

Strategic wins : pricing decisions that built developer trust

1. Pricing the database like serverless compute

By billing Redis per command ($0.2 / 100K) with true scale-to-zero, Upstash matched its pricing to the mental model developers already had for AWS Lambda and edge functions. This removed the single biggest objection to managed Redis for serverless apps — paying for an idle instance — and aligned cost directly with value delivered. See why this alignment matters in choosing the right usage metric and the broader shift away from per-seat licensing.

2. A budget cap that defuses the bill-shock objection

The hard monthly budget cap on Pay as You Go turns “usage-based pricing might blow up my bill” into a non-issue: the bill is provably bounded. This is a textbook answer to the cost-unpredictability problem that makes finance teams nervous about consumption pricing, and it lets developers adopt without procurement friction.

3. Unbundling production guarantees into a flat add-on

Selling the SLA, multi-zone HA, SOC-2, and encryption at rest as a $200/mo Prod Pack — available on any paid plan without a sales call — lets small teams buy enterprise-grade reliability self-serve. It also creates a clean, predictable expansion-revenue lever that doesn’t require renegotiation, a pattern worth studying alongside how to package usage-based tiers.

4. Lowering the floor instead of raising the ceiling

The March 2025 repricing cut the Redis entry flat-rate from $280/mo to $10/mo rather than hiking per-request rates. Lowering the floor expanded the addressable base of small production workloads while keeping the usage meter — and therefore expansion revenue from scaling customers — completely intact. This is a rare example of a pricing change that improves both top-of-funnel conversion and net revenue retention.

Areas to improve : where the model adds friction or risk

1. Make per-region replication cost impossible to miss

The biggest source of surprise on a global Redis bill is that writes are re-counted per read region — doubling or tripling the effective command cost. While the FAQ documents it, the headline “$0.2 / 100K” framing invites under-budgeting. A live multiplier in the pricing UI (region count × write share) and a prominent example in the hero would prevent the gap. See common usage-billing pitfalls.

2. Publish an Enterprise price anchor

Enterprise is “Custom” across every product with no published starting point. A “from $X/mo” anchor or a representative example bill would help mid-market buyers self-qualify before a sales call and reduce the opacity that contrasts with Upstash’s otherwise unusually transparent pricing. Related: why transparent pricing accelerates the funnel.

3. Provide a unified cross-product cost estimator

With five products on five different units (commands, requests, messages, steps, queries) plus storage and bandwidth, modeling a multi-product bill is genuinely hard. A single calculator that sums all five products — rather than per-product pages — would shorten the path from evaluation to commitment. See how to communicate usage pricing clearly.

4. Commit to a product-longevity policy after the Kafka sunset

Deprecating Serverless Kafka in 2024 — even with a six-month wind-down — signals to buyers that any product line can be retired. A published support-lifecycle or migration-guarantee policy would reassure teams building on newer products like Workflow and Search.

Key takeaways

1. Match the pricing unit to the buyer’s mental model. Upstash priced a stateful database per operation because its serverless-native buyers already think in per-invocation terms. Picking the metric the customer already uses to reason about value removes adoption friction before it starts.
2. A hard budget cap is the cleanest answer to bill-shock fear. Making the bill mathematically bounded — not just alerted on — converts the strongest objection to usage pricing into a selling point. Most usage-metered vendors only offer alerts; a true cap is differentiating.
3. Unbundle reliability instead of gating it. Selling production guarantees as a flat self-serve add-on lets small customers buy enterprise-grade features and creates predictable expansion revenue without renegotiation.
4. Lower the floor, keep the meter. Cutting the entry flat-rate price 28× while leaving per-request rates untouched grew the funnel without sacrificing expansion revenue — a move other infrastructure vendors should study before reflexively raising prices.
5. Document the cost amplifiers honestly — but surface them louder. Upstash’s FAQ openly explains retries-as-messages and per-region replication, which builds trust, but burying multipliers below a clean headline rate still invites under-budgeting. Transparency in docs is necessary but not sufficient; the multiplier belongs in the pricing UI.

UBP implications

1. Scale-to-zero is the killer feature of pure-usage data pricing. When idle cost is genuinely $0, usage-based billing becomes strictly better than provisioned pricing for spiky and long-tail workloads — the model wins on the economics, not just the philosophy.
2. Bounded usage pricing (cap + meter) is a distinct, adoptable category. Pairing a per-request meter with a hard budget cap gives buyers the upside of consumption pricing without the downside risk, and may be the template for usage pricing in cost-sensitive segments.
3. Add-on packaging beats tier-gating for reliability features. Unbundling SLA/HA/compliance into a flat add-on monetizes production readiness across the whole base rather than only the enterprise tier, smoothing expansion revenue and widening who can buy it.

Sources

• Upstash pricing hub (accessed 2026-06-03)
• Upstash Redis pricing (accessed 2026-06-03)
• Upstash Vector pricing (accessed 2026-06-03)
• Upstash QStash pricing (accessed 2026-06-03)
• Upstash Workflow pricing (accessed 2026-06-03)
• Upstash Search pricing (accessed 2026-06-03)
• Upstash Redis pricing & limits docs (accessed 2026-06-03)
• Upstash Vector pricing docs — second-source validation (accessed 2026-06-03)
• Upstash QStash pricing docs — second-source validation (accessed 2026-06-03)
• Upstash blog — New Pricing and Increased Limits for Upstash Redis (accessed 2026-06-03)
• Upstash blog — Announcing Upstash Workflow and Deprecating Upstash Kafka (accessed 2026-06-03)
• Upstash blog — QStash: Messaging for the Serverless (accessed 2026-06-03)
• Upstash blog — Introducing Upstash Vector (accessed 2026-06-03)

Bottom line

Upstash is the cleanest expression of a simple thesis: serverless data should bill like serverless compute — per operation, scaling to zero, with a hard budget cap so the bill can never surprise you. Across five products it has held its per-request rates steady for years, lowered its entry price rather than raised it, and sold production reliability as a flat add-on instead of an enterprise gate. The friction that remains is multiplier opacity (per-region replication, retries) and a fully custom Enterprise tier — but for serverless and edge developers, Upstash is among the most honestly-priced infrastructure on the market.

Compare Upstash against other infrastructure and data-platform pricing models in the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Current snapshot — per-request rates across five products

Jun 2026

Redis $0.2/100K commands, Vector $0.4/100K requests, QStash & Workflow $1/100K (messages/steps), Search $0.05/1K requests (Preview). Fixed plans: Redis $10–$1,500/mo, Vector $60/mo, QStash/Workflow $180 (1M) and $420 (10M). Prod Pack production add-on +$200/mo per database; Enterprise custom.

QStash and Workflow reach General Availability

Oct 2025

QStash and Workflow are declared production-ready / GA. Fixed plans on each settle at $180/mo (1M/day) and $420/mo (10M/day), with Pay-as-You-Go at $1 per 100K messages or steps. (upstash.com/blog/qstash-qa)

Redis repriced — Fixed plans from $10/mo replace $280+ Pro plans

Mar 2025

Upstash replaces the old Redis 'Pro 2K' ($280/mo) and 'Pro 10K' ($680/mo) plans with new Fixed plans starting at $10/mo (up to $1,500/mo), expands the free tier from 10K commands/day to 500K commands/month, and lifts Pay-as-You-Go storage from 10GB to 100GB with 200GB free bandwidth. Wayback confirms the $280/$680 Pro model on the Sept 2023 snapshot. (upstash.com/blog/redis-new-pricing)

Workflow launches; Serverless Kafka deprecated

Sep 2024

Upstash Workflow — a durable execution engine built on QStash — launches at $1 per 100K steps, and Upstash announces Serverless Kafka will be discontinued over the following six months to focus on QStash and Workflow. (upstash.com/blog/workflow-kafka)

Upstash Vector launches — serverless vector database

Jan 2024

Vector launches for AI/LLM workloads at $0.4 per 100K requests plus $0.25/GB storage, with a free tier of 10K daily query/update. Wayback confirms a Vector tab appearing on upstash.com/pricing by Feb 2024. (upstash.com/blog/introducing-vector-database)

QStash launches — serverless message queue / scheduler

Jul 2022

QStash launches as a message queue and task scheduler for serverless runtimes, priced $1 per 100K messages where every delivery attempt (including retries) counts as one message. (upstash.com/blog/qstash-announcement)

Serverless Redis launches with per-request pricing

Mar 2021

Upstash launches managed serverless Redis billed at $0.2 per 100K commands with a free tier and budget cap — pricing the database by the operation rather than the provisioned instance, so it scales to zero with no idle cost.