We introduced QPS limits to Developer Plans as a way to prevent abuse. They did that. They also did something we didn't want: they put an artificial ceiling on what our users could do.
The most consistent piece of feedback we've heard from developers on these plans has been some version of "QPS limits are the worst." We agree.
Today we're removing them. Billing on Developer Plans is now based on actual compute usage, measured in vCPU-seconds. Every plan includes built-in burst capacity. Pricing on the base plans hasn't changed.
Why we're changing this
Some queries take 20 ms of CPU. Or less. Capping the user running those queries at 10 or 15 QPS made no sense. They had compute headroom. We were the ones telling them to stop.
QPS is a proxy. What we actually care about, and what actually affects the platform, is compute. So we're billing on the thing that matters and dropping the proxy.
The new model is meant to be fair in both directions: you pay for the compute you actually use, and the signal to upgrade is when you've genuinely outgrown your plan's capacity, not when you've hit an arbitrary request count.
What changed
QPS limits and QPS overages removed. Your plan no longer caps how many queries per second you can run. The old "active minutes" metric is also gone.
Billing based on vCPU-seconds. Instead of counting requests, we measure the actual compute each query consumes. A lightweight query that takes 0.01 vCPU-seconds costs less than a heavy one that takes 0.5. You pay for what you compute, not how often you call the API. The overage rate is $0.0002 per vCPU-second above your plan's baseline, the same across all Developer Plans.
Built-in burst capacity. Every plan includes a baseline vCPU allocation plus a burst multiplier. S-1/4 through S-2 plans can burst up to 3x their baseline. S-4 and S-8 can burst up to 2x. During short bursts, queries run uninterrupted. You're billed for usage above baseline at $0.0002/vCPU-sec. Under sustained load beyond your plan's burst ceiling, read queries may be throttled. Ingestion is never blocked.
Ingestion is never throttled. Data ingestion and batch operations are never blocked by plan capacity, regardless of compute load. Read-query throttling under sustained heavy bursts replaces the old QPS cap with a limit tied to what your plan can actually process. Check the limits documentation for specifics on concurrency and burst duration per plan.
Flexible plan changes. You can upgrade or downgrade at any time. The previous 24-hour lock between plan changes is gone. If you change plans within the first 24 hours, a minimum fee equivalent to one day of the former plan applies.
CPU On Demand. Batch operations like populates and copies run on on-demand compute and are billed separately. This is independent of your plan's baseline compute. See the pricing documentation for rates.
What this looks like in practice
Say you're on an S-1 plan ($99/mo). Your baseline is 1 vCPU-second/sec with 3x burst capacity.
Your traffic spikes and your queries average 2 vCPU-seconds/sec for 10 minutes. You're within your burst ceiling (under 3 vCPU-sec/sec), so nothing gets throttled. But you used 1 vCPU-second/sec above your baseline for 600 seconds. That's 600 vCPU-seconds of overage.
Cost of that spike: 600 x $0.0002 = $0.12.
Plan updates
| Plan | vCPUs (baseline) | Burst | Price |
|---|---|---|---|
| S-1/4 | 0.25 | 3x | $25/mo |
| S-1/2 | 0.5 | 3x | $49/mo |
| S-1 | 1.0 | 3x | $99/mo |
| S-2 | 2.0 | 3x | $199/mo |
| S-4 | 4.0 | 2x | $399/mo |
| S-8 (new) | 8.0 | 2x | $799/mo |
All plans include 25 GB of storage ($0.058 per additional GB).
The S-3 plan is no longer available to new customers. Existing S-3 customers can stay on it.
Base plan pricing is unchanged. The monthly fee for each tier is the same as before.
How the rollout works
We've been running this model with a group of customers since April. The feedback shaped the final design.
Changes take effect when your current billing cycle ends. Your existing billing period is not interrupted. Free plan users will also move to the vCPU-based structure.
Learn more
For the full breakdown of how vCPU billing works, including burst mode, overages, and on-demand compute, check the pricing documentation.
You can monitor your vCPU consumption in your organization's consumption dashboard.
If you have questions about how the change affects your workload, reach out in our Community Slack or contact support.