Credits
Last checked:Cerebrium free tier: Free plan, compute billed
creditsserverlessgpucerebrium
A serverless GPU platform for deploying AI apps in Python, with a Hobby plan whose $0 price covers the plan fee, not compute. Official site
Cerebrium free tier, the facts
| Amount | Free plan, compute billed |
|---|---|
| Grant type | Monthly free tier |
| Expiry | No credit exists to expire; the Hobby plan itself has no stated time limit |
| Payment method required | Not stated officially |
| Withdrawable as cash | No |
| What I ran on it | Not tested yet |
Filed under: GPU cloud free creditsServerless free tiers and credits
Good for
- Deploying serverless GPU apps with no monthly platform fee
- Small projects that fit 3 apps and 5 concurrent GPUs on the Hobby plan
Not for
- Free GPU time: every compute second is billed from the start
- Anyone expecting the circulating $10 or $30 signup credit: no official page states one
Cerebrium’s Hobby plan is priced “Free + compute / month”: the plan fee is $0, and every second of compute is billed at per-second rates. There is no dollar signup credit on the official pricing page as of 2026-09-02, no card requirement is stated, and the only free resource quantified officially is the first 100GB of storage. If you came here expecting free GPU credit, the accurate summary is: the plan is free, the compute is not.
The circulating $10 and $30 claims, corrected
Directories, YC-adjacent posts, and older blog writeups attach “$10 free credit” or “$30 free credit” to Cerebrium. As of 2026-08-21 neither figure appears on the official pricing page or terms of service. This matters for planning: on Modal, $30 per month of real credit means your first GPU hours cost nothing. On Cerebrium, your first GPU second is billed, and since 2026-09-09 so is Beam’s, after Beam withdrew the monthly credit it used to publish. Do not budget around a credit that no official page states.
What the free plan actually includes
The Hobby plan caps you at 3 seats, 3 deployed apps, and 500 containers with 5 concurrent GPUs. The first 100GB of storage is free. Log retention is listed as 1 day, though another part of the official site says 7 days, a contradiction I am flagging rather than resolving. The paid Standard plan is $100 per month plus compute.
Compute rates, for scale
Official per-second rates include a T4 at $0.000164 per second, an L4 at $0.000222, an A100 80GB at $0.000583, and an H100 at $0.000944 per second. Roughly $3.40 per H100 hour, all of it billed to you from the first second on the Hobby plan.
One more credit gotcha
Per Cerebrium’s terms, AWS and GCP credits cannot be applied to Cerebrium. If your plan was to burn a Google Cloud $300 trial credit through a serverless layer, that does not work here.
What I have done with it
Not tested yet. I have not deployed on Cerebrium for this entry, so it reports only what the official pages state.
Choose Cerebrium if you want its deployment workflow and are ready to pay for compute from day one. Skip it if you wanted a free-credit runway: Modal gives $30 per month of actual credit on an otherwise similar serverless GPU platform, and RunPod is the equivalent no-credit reality check for rented pods. Beam matched Modal until 2026-09-09 and now sits in the same class as Cerebrium.
What changed here
Added Cerebrium: Free plan, compute billed.
Plus 1 recheck that found nothing changed. Every row links to the full record on the changelog, with the official page it was checked against.
Related entries
Modal: $30 per month
Beam: None
RunPod: None
Sources
Last checked:All numbers on this page come from the official pages above, checked on the date shown. If a figure is not stated there, this page says "Not stated officially" instead of guessing.