Database pricing, compared

TursoDeveloper

Monthly-billing prices (the page shows annual by default — $4.99, $24.92 and $416.58/month billed yearly). Turso has no compute-hour or egress charge; it bills rows read and written instead, modelled on reads_m / writes_m: Free includes 500M rows read / 10M written (hard limit), Developer 2.5B / 25M ($1 per extra billion read = $0.001 per million, $1 per extra million written), Scaler 100B / 100M ($0.80 per billion read = $0.0008 per million, $0.80 per million written). Rows read counts every row scanned, not just rows returned. Embedded-replica syncs (3 / 10 / 24 GB included, then $0.35 / $0.25 per GB) are not modelled. Pro ($499/month) is left out as it's never the cheaper choice at these sizes.

PlanetScalePostgres PS-5 (single node)

AWS us-east-1 prices. PS-5 is 1/16 vCPU and 512 MiB RAM; bigger sizes (PS-10 $10 single / $30 HA on arm64, and up) cost more. Each cluster includes the first 10 GB of network-attached storage and production HA branches include 100 GB/month egress, a single-node PS-5 only 10 GB; extra storage is $0.125/GB and egress $0.06/GB (from PlanetScale's published storage and egress rate endpoints for us-east). Compute hours beyond 730 are a second cluster at the same monthly rate prorated per hour. Backups beyond the included amount and Vitess (MySQL, from $39 for PS-10) are not modelled. No free tier.

FirebaseBlaze (pay as you go)

Prices are for Iowa (us-central1); multi-region locations such as nam5 cost more per operation. Storage is listed at $0.000205479 per GiB-hour ≈ $0.15 per GiB-month. Daily quotas are converted × 30 and assume steady traffic. Deletes ($0.01 per 100,000, 20,000 a day free) aren't modelled. Egress to internet destinations is graduated: free to 10 GiB, $0.12 to 1 TiB, $0.11 to 10 TiB, $0.08 beyond — the first paid tier is modelled; traffic between US Google Cloud regions is free and China/Australia destinations cost more. Firestore is serverless, so compute_hours doesn't apply. Committed-use discounts (1 or 3 years, 20–40% off operations) are left out. Spark is Firebase's no-billing plan, capped at the same free quota.

NeonLaunch

Compute hours are priced at the smallest compute size, 0.25 CU (1 GB RAM) — Launch charges $0.106 per CU-hour, so $0.0265 per hour at 0.25 CU; larger or autoscaled computes cost proportionally more. Free includes 100 CU-hours, 1 GB of storage and 5 GB of egress per project (400 hours at 0.25 CU), for up to 100 projects, and suspends compute when the hours run out. Scale plan ($0.222/CU-hour, same $0.35/GB-month storage) adds more features but always costs more, so it's left out. No monthly minimum on Launch.

SupabasePro

Pro includes $10/month of compute credits, which cover one project on Micro compute (shared CPU, 1 GB RAM, $0.01344/hour ≈ $10/month) — hours beyond 730 are a second Micro project. Larger compute sizes cost more per hour. The Pro fee also covers auth and file storage. Egress is uncached egress; cached egress has its own 250 GB allowance and costs $0.03/GB beyond it. Free (Nano compute, shared CPU, 500 MB RAM) allows 2 active projects, which pause after a week of inactivity.

ConvexProfessional

Professional is $25 per developer per month (one developer shown). Convex bills function calls, not rows: the model counts each read as one query call and each write as one mutation call ($2.20 / $2 per million), and puts the whole included allowance (1M calls on Free/Starter, 25M on Professional) on reads, although Convex pools it across all calls. A query that reads many documents is still one call, so read-heavy apps cost less than shown. Egress is 1 GB / 50 GB included, then $0.132 / $0.12 per GB. There's no compute-hour price (query and mutation compute is free on these plans). Left out: database I/O (1 GB / 50 GB included, then $0.22 / $0.20 per GB), action compute (20 / 250 GB-hours, then $0.33 / $0.30 per GB-hour), file and search storage. Free stays free within its allowances; Starter is the same plan with pay-as-you-go overage enabled. US-region prices.

MongoDBAtlas Dedicated M10

M0 caps data transfer at 10 GB out per rolling seven days (about 43 GB a month) and throughput at 100 operations per second; the model applies the transfer cap. Flex starts at $0.011/hour and rises with operations, capped at $30/month — the model shows the starting rate, which covers up to 100 ops/second (about 260M operations a month if spread evenly). Atlas doesn't bill per operation, so reads and writes aren't priced. The page also lists shared M2 ($9/month, 2 GB) and M5 ($25/month, 5 GB) tiers, not modelled. M10 (2 GB RAM, 2 vCPUs) is $0.08/hour (from $56.94/month) with 10 GB storage in the base price; larger disks (up to 128 GB) cost extra and aren't priced on the page, so the model caps M10 at 10 GB. Free and Flex don't charge egress; Dedicated clusters pay the cloud provider's egress rates (not modelled). Backups and add-ons (Search, Vector Search) not modelled.