# Cerebrium

Official pricing: https://cerebrium.ai/pricing  
Category: gpu-cloud · Isolation: container

## Pricing regimes (raw)

- **Interruptible** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot, alt]
- **Protected compute** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [alt]
- **B200, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]
- **H200, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]
- **H100, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]
- **RTX-PRO-6000, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]
- **A100-80GB, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]
- **A100-40GB, serverless** (resource): $0.02358/vCPU-h, $0.007992/GiB-h [spot]

## Features

Yes: snapshots, memory snapshots, persistent disk, volumes, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, HTTPS preview URLs, custom domains, private networking, SOC 2, HIPAA, GPU, Python, wake on request, webhooks, audit logs, EU data residency, snapshots on demand, inbound access rules, extra volumes, shared volumes

No: fork/clone, pause/resume, full VM (own kernel), Docker inside, nested virtualization, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, SSH, public IPv4, egress allowlist, static egress IP, self-hosting, BYOC, open source, SSO, arm64, Windows, macOS, Node.js, HTTP method/path egress rules, live resize, live fork (no pause), memory fork, MCP server, automatic snapshots, agent harness API, hosted agent API (their own agent), gVisor or VM (no shared kernel), firewall inside (nftables), secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/cerebrium.json → feature_evidence

## Caveats

- CPU $0.00000655/vCPU-s, RAM $0.00000222/GB-s. Hardware docs say actual CPU/memory usage; pricing FAQ also says resources allocated while processing. Record utilization ambiguity rather than promise savings.
- Default interruptible tier can be preempted or consolidated. Protected tier costs 2× all compute including GPU, CPU and RAM; storage unaffected.
- Storage first 100 GB free then $0.05/GB-month. Standard $100 is plus compute (no stated credits). Pricing comparison conflicts on Standard seats and Hobby log retention.
- GPU listed per-second rates converted without rounding to marketing hourly approximations.
- Null values mean unknown, not free/unlimited. No model-token cost, tax or unverified ancillary charge included.
- Native browser/device/task/credit prices with unknown CPU/RAM are deliberately non-priceable. Check regime before interpreting any partial resource estimate.
- GPU modes grafted from the v4-gpu research card (per-GPU rates; exact rows in research/gpu.json).
- GPU concurrency per plan (pricing page 2026-09-28): Hobby 5 GPUs, Standard ($100/mo) 30 GPUs, Enterprise unlimited; plan.concurrency holds the CPU container limits (500 / 1,000). GPU cap not enforced by the engine.
- Per-container max (docs/hardware/cpu-and-memory): H100/H200/B200 24 vCPU / 256 GB; A100 12 vCPU / 140 GB; L4/T4 48 vCPU / 192 GB; L40S/A10/RTX PRO 6000 not listed.
- AWS Inferentia2 (INF2, Hobby+) and Trainium (TRN1, Enterprise) are offered but have no public price; TRN1 containers up to 128 vCPU / 512 GB. Max 8 GPUs per container for every GPU type.
- Builds are billed at the app's compute rate when requirements/code change (cached steps cheaper). min_replicas>0 bills 24/7.
- Memory+GPU checkpointing (snapshots for cold start) is early beta; no separate storage price published.
- Platform cold-start time is not billed; user initialization code outside the request function (e.g. loading a model into GPU memory) is billed.
- Persistent storage: one volume per region per project (mounted /persistent-storage), resizable 50 GB-1 TB self-serve; global apps also get a fixed 1 TB /global-persistent-storage volume. Pricing page: first 100 GB free then $0.05/GB-mo; whether billing is on provisioned size or used bytes is not stated.

## How this provider charges


YC: Winter 2022; directory status **Active**. Research scope: public-product-researched. Source: https://www.ycombinator.com/companies/cerebrium

Serverless Infrastructure Platform for AI

## Regime table

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| Interruptible | spot, alt | resource; CPU alloc, memory alloc | $0.02358/vCPU-h + $0.007992/GiB-h; multiplier 1 | https://cerebrium.ai/pricing |
| Protected compute | alt | resource; CPU alloc, memory alloc | $0.02358/vCPU-h + $0.007992/GiB-h; multiplier 2 | https://cerebrium.ai/pricing |
| Hobby | Plan limits apply | Monthly fee; credits only as stated | fee=0; included usage=$0; concurrent=500; max session h=None | https://cerebrium.ai/pricing |
| Standard | Plan limits apply | Monthly fee; credits only as stated | fee=100; included usage=$0; concurrent=1000; max session h=None | https://cerebrium.ai/pricing |
| Enterprise | Plan limits apply | Monthly fee; credits only as stated | fee=None; included usage=$None; concurrent=None; max session h=None | https://cerebrium.ai/pricing |
| Retained disk / GiB-month | Lifecycle/scope must be checked | Native add-on | $0.05 | https://cerebrium.ai/pricing |
| Snapshots / GiB-month | Lifecycle/scope must be checked | Native add-on | unverified, not $0 | https://cerebrium.ai/pricing |
| Egress / GiB | Lifecycle/scope must be checked | Native add-on | unverified, not $0 | https://cerebrium.ai/pricing |
| IPv4 / month | Lifecycle/scope must be checked | Native add-on | unverified, not $0 | https://cerebrium.ai/pricing |

## Verified billing facts and gotchas

1. CPU $0.00000655/vCPU-s, RAM $0.00000222/GB-s. Hardware docs say actual CPU/memory usage; pricing FAQ also says resources allocated while processing. Record utilization ambiguity rather than promise savings.
2. Default interruptible tier can be preempted or consolidated. Protected tier costs 2× all compute including GPU, CPU and RAM; storage unaffected.
3. Storage first 100 GB free then $0.05/GB-month. Standard $100 is plus compute (no stated credits). Pricing comparison conflicts on Standard seats and Hobby log retention.
4. GPU listed per-second rates converted without rounding to marketing hourly approximations.

Native pricing (not coerced into incompatible engine units):
```json
{
  "storage_free_account_gb": 100
}
```

## Worked example

Requested: 4 vCPU/8 GiB, 50 simultaneous × 8 h/day × 22 days = **8,800 instance-hours**, 30% CPU; 50 GiB snapshots and 100 GiB egress. A month is 730 hours only when explicitly used in rate conversion.

Conservative allocated-resource scenario: 8,800×(4×0.02358+8×0.007992)=$1,392.65 interruptible, or $2,785.31 protected. If CPU is truly utilization-metered, 30% CPU/full RAM gives $811.64/$1,623.29. Storage 50 GB fits the 100 GB persistent-volume allowance only if it is eligible volume data, not automatically snapshot storage. Egress and snapshot charges remain unknown.

## Sources and coverage

- https://www.ycombinator.com/companies/cerebrium
- https://cerebrium.ai/pricing
- https://cerebrium.ai/docs/hardware/cpu-and-memory
- https://cerebrium.ai/docs/llms-full.txt
- https://cerebrium.ai/
- https://cerebrium.ai/docs/getting-started/introduction
- https://cerebrium.ai/llms.txt

Page captures and documentation indexes/bundles are in `raw/`; provider JSON lists the evidence paths. HN query results were captured separately as `<id>--hn.txt`; they are discovery leads, not current tariff authority. Downloaded docs do not imply every feature was verified; unsupported assertions remain null.