MongoDB Atlas has three tiers: Free (M0, 512MB, permanently free), Flex ($8-30/month, capped usage-based pricing that replaced the old Serverless and M2/M5 tiers in February 2025), and Dedicated (from roughly $0.08/hour, about $58/month for the smallest M10 cluster, billed continuously regardless of traffic). Enterprise Advanced, for self-managed deployments, is custom-quoted.
- Flex replaced the old Serverless and shared M2/M5 tiers in a February 2025 consolidation
- Dedicated clusters bill 24/7 based on provisioned capacity — not actual usage or traffic
- Search and Vector Search nodes are metered separately from core database compute
- The jump from Flex's $30 cap to Dedicated's continuous billing is the single biggest cost step in the product line
MongoDB Atlas simplified its lower tiers in February 2025, folding the old Serverless instances and shared M2/M5 clusters into a single Flex tier with a predictable cap. That's a genuine improvement in predictability at the low end. The tradeoff shows up once an application outgrows Flex and needs a Dedicated cluster — a fundamentally different billing model that charges for provisioned infrastructure around the clock, not for what actually gets used.
This guide covers what changed in the 2025 tier consolidation, what a Dedicated cluster actually costs once you're paying for uptime rather than usage, and where Search and Vector Search add their own separate charges.
1. How much does MongoDB Atlas cost?
Source: mongodb.com/pricing (official), verified Sep 19, 2026.
Flex pricing is usage-based but capped between $8 and $30 a month, giving small applications a predictable ceiling. Dedicated clusters bill by the hour based on provisioned instance size (RAM, vCPU, storage) continuously — 24 hours a day, 7 days a week — regardless of whether the database is actively serving traffic. There's no way to pause billing on a Dedicated cluster without actually deleting it.
2. What each plan includes
| Feature | Free (M0) | Flex | Dedicated (M10+) | Enterprise Advanced |
|---|---|---|---|---|
| Automatic backups | ✕ | ✓ | ✓ | ✓ |
| Dedicated compute & storage (no sharing) | ✕ | ✕ | ✓ | ✓ |
| Atlas Search included | ✕ | ✕ | ✕ | ✕ |
| Multi-region & advanced sharding | ✕ | ✕ | ✓ | ✓ |
| Self-managed / on-prem deployment | ✕ | ✕ | ✕ | ✓ |
3. Cost calculator
4. Hidden costs to budget for
| Scenario | Annual cost | Change |
|---|---|---|
| Flex, light usage | ~$96-360/yr | Capped, predictable |
| Dedicated M10, 24/7 billing | ~$696/yr | Bills continuously regardless of traffic |
| Atlas Search or Vector Search added | Variable, separate meter | Not included in base compute pricing |
Atlas Search and Vector Search run on their own dedicated search nodes, billed separately from the core database cluster's compute and storage — an application that adds full-text or vector search capability should expect that as a distinct new line item, not a feature bundled into the existing cluster cost. Combined with 24/7 Dedicated cluster billing, a database that's provisioned generously for occasional peak traffic but runs quiet most of the time can end up costing considerably more than its actual utilization would suggest.
5. How MongoDB Atlas pricing has changed
6. Which plan should you choose?
- For learning, prototyping, or very small applications, Free (M0)'s 512MB is permanently available at no cost.
- For production applications with modest, somewhat unpredictable traffic, Flex's capped pricing avoids surprise bills as usage fluctuates.
- Once an application needs dedicated resources not shared with other tenants, or consistent performance guarantees, Dedicated is the right move — but budget for continuous billing, not usage-based billing.
- Organizations needing to self-manage MongoDB rather than use Atlas's cloud service should evaluate Enterprise Advanced separately.
7. Is it worth the price?
For applications with steady, predictable traffic that genuinely needs dedicated resources, MongoDB Atlas's Dedicated tier pricing is competitive against comparable managed database services. The friction point is specifically applications with spiky or intermittent traffic patterns provisioned onto Dedicated infrastructure sized for their peak — that combination pays for capacity that sits unused a meaningful share of the time, which is a real cost inefficiency worth addressing through right-sizing rather than just accepting.
8. How to pay less
- Stay on Flex as long as its capped pricing genuinely covers your traffic patterns — don't move to Dedicated preemptively based on anticipated growth that hasn't materialized yet.
- Right-size Dedicated cluster instances to actual sustained load, not peak load — since billing is continuous, provisioning for rare peaks means paying that rate every hour, not just during peaks.
- Review Atlas Search and Vector Search usage separately from core database costs — these are easy to overlook as a distinct, growing line item.
- Consider auto-scaling configurations where available to reduce continuous over-provisioning during predictably quiet periods.
For teams evaluating whether Dedicated is justified yet, a rough guideline is that Flex's $30 cap corresponds to meaningfully more headroom than the old shared M2/M5 tiers offered — many applications can stay on Flex longer than they might assume before genuinely needing Dedicated infrastructure.
9. Alternatives and what they cost
| Tool | Entry price | Cheaper when |
|---|---|---|
| PlanetScale (MySQL) | Free / usage-based | MySQL compatibility fits your stack better than MongoDB's document model |
| Supabase | Free / $25/mo (Pro) | Postgres-based backend with bundled auth and storage fits your needs |
| Self-hosted MongoDB (Community) | Infrastructure cost only | You have the operational capacity to manage your own database infrastructure |
10. The bigger picture: why MongoDB Atlas bills drift
Continuous, capacity-based billing that doesn't automatically scale down with actual usage is a pattern across most managed infrastructure services, not just MongoDB — and it rewards exactly the kind of ongoing utilization review that's easy to deprioritize once a system is running smoothly. Zapro AI is built to flag this kind of infrastructure spend pattern across your vendor list.
11. The bottom line
MongoDB Atlas's 2025 tier consolidation made small-application pricing genuinely simpler and more predictable. The real cost discipline required is at the Dedicated tier, where billing runs continuously against provisioned capacity rather than actual usage — right-sizing that provisioning to real, sustained load (not peak load) is the single biggest lever for keeping the bill in line with what you're actually using.
FAQs
It was consolidated, along with the shared M2/M5 tiers, into a single Flex tier in February 2025 — Flex offers capped, predictable pricing between $8 and $30/month for similar small-application use cases.
On Dedicated clusters, yes — billing is continuous based on provisioned instance size, regardless of whether the database is actively handling traffic. Flex's usage-based, capped pricing behaves differently and scales more closely with actual activity.
No. Atlas Search and Vector Search run on separate, dedicated search nodes with their own billing, distinct from core database compute and storage costs.
The smallest Dedicated tier (M10) starts around $58/month based on roughly $0.08/hour of continuous billing, with cost scaling up based on RAM, vCPU, and storage as you choose larger instance sizes.
The most common way is a Dedicated cluster provisioned for peak traffic but running well below that most of the time, or Atlas Search costs accumulating unnoticed. A vendor spend platform that compares provisioned capacity against actual utilization is the most reliable way to catch this.