# Cloud Run GPUs (slim)

> Cloud Run GPUs attach one NVIDIA L4 or RTX PRO 6000 Blackwell GPU to a Cloud Run service, job or worker pool, so a container serves models serverlessly, billed per second and able to scale to zero.

- Full: https://www.anchorterminal.com/tools/cloud-run-gpus.md (~7,600 tokens) · this version ~1,480 tokens · JSON https://www.anchorterminal.com/tools/cloud-run-gpus.json · canonical https://www.anchorterminal.com/tools/cloud-run-gpus
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-07 · pricing https://www.anchorterminal.com/tools/cloud-run-gpus#pricing · reviews https://www.anchorterminal.com/tools/cloud-run-gpus#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**B · 68.4/100 · rank #288 of 1486 · #2 in GPU & serverless compute · not agent-ready · confidence medium**


Assessment: Serverless GPUs on Cloud Run, an L4 or RTX PRO 6000 per instance, billed per second, scaling to zero and starting in about 5 seconds before the model loads, with a published SLA and IAM. One GPU per instance, 3 GPUs of default quota a region and instance-based billing limit it to small and mid-size models.

## Facts

- Kind: Platform · vendor: Google Cloud · category: GPU & serverless compute · legal entity: Google LLC · provenance 99/100
- Endpoint: `https://run.googleapis.com` ()
- Auth: OAuth · pricing: Pay per use · x402: no · licence: Proprietary service under the Google Cloud Platform terms. Client libraries are Apache-2.0
- Benchmark probes: not run yet, so Performance is pending
- Live, from our pollers (not part of the score): up at 2026-11-07 17:49 UTC · uptime 30 days 100.0% (195 probes) · p50 35 ms · p95 110 ms over 24 hours
- Vendor status page: not read, no machine-readable status found (2026-11-07 04:47 UTC)
- Surface listed: GPUs on Cloud Run services, jobs and worker pools, configured through gcloud, YAML, Terraform or the Cloud Run Admin API, read from docs.cloud.google.com (GPU pages for services and jobs, GPU best practices, quotas, release notes), the pricing page, the SLA and status.cloud.google.com on 6 November 2026. Nothing was deployed
- Scope: The GPU option of Cloud Run. Google Cloud's other model and compute products are separate, and no Anchor Terminal listing covers Cloud Run itself
- GPUs: One GPU per instance, NVIDIA L4 (24 GB) or RTX PRO 6000 Blackwell (96 GB), attached to one container
- Regions: L4 in regions including europe-west1, europe-west4, asia-southeast1 and asia-south1 (by invitation); RTX PRO 6000 in us-central1, europe-west4, asia-southeast1 and asia-south2, among others listed on the GPU page
- Zonal redundancy: On by default for new services, reserving GPU capacity across zones for failover at a higher price; off is cheaper and fails over on a best-effort basis
- Billing: Instance-based billing only, for the whole instance lifecycle. Minimum instances bill at the full rate while idle. No per-request fees
- Model loading: The best-practices page recommends downloading models from Cloud Storage at startup in parallel, and keeping models in the image only below about 10 GB
- Prices: NVIDIA L4, no zonal redundancy $0.67 per GPU-hour; NVIDIA L4, zonal redundancy $1.05 per GPU-hour; NVIDIA RTX PRO 6000, no zonal redundancy $1.31 per GPU-hour; NVIDIA RTX PRO 6000, zonal redundancy $2.05 per GPU-hour

- Scores: Reliability 70, Performance pending, Schema & documentation 81, Agent ergonomics 66, Security & auth 78, Payments & pricing 20, Task success pending, Maintenance & community 80, Transparency & trust 84 · total over the 7 assessed categories
- Why: Reliability, Hosted reading. · Schema & documentation, API reading. · Agent ergonomics, Platform reading. · Security & auth, OAuth 2.0 tokens from service accounts under IAM. · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, Closed service with official SDKs. · Transparency & trust, Closed service under the Google Cloud terms, with Apache-2.0 client libraries (15).
- Sources: 9, open questions: 3, both in the full twin
- Capabilities: compute.serverless, compute.gpu, compute.endpoints, compute.containers, compute.batch
- JSON: https://www.anchorterminal.com/api/v1/tools/cloud-run-gpus.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/cloud-run-gpus.svg` or a link to https://www.anchorterminal.com/tools/cloud-run-gpus from a page on google.com or one of its subdomains, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.
- Support: https://cloud.google.com/support-hub; https://support.google.com/; more in the full twin

## Before you call it

1. Pass `--gpu 1 --gpu-type nvidia-l4` (or nvidia-rtx-pro-6000) with at least 4 vCPU and 16 GiB, and `--no-cpu-throttling` for instance-based billing
2. Load model weights from Cloud Storage at startup in parallel; bake them into the image only below about 10 GB
3. Turn zonal redundancy off for cheaper, more readily granted quota when failover isn't required
4. Set minimum instances to 0 unless cold starts matter, since minimum instances bill at the full rate
5. Send OAuth tokens in the `Authorization` header, never as `access_token` in the URL

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Modal | B | 62.7 | compute.gpu, compute.serverless, compute.endpoints, compute.batch, compute.containers | https://www.anchorterminal.com/tools/modal.min.md |
| Beam | C | 56.2 | compute.gpu, compute.serverless, compute.endpoints, compute.batch, compute.containers | https://www.anchorterminal.com/tools/beam.min.md |
| Runpod | C | 55.6 | compute.gpu, compute.serverless, compute.endpoints, compute.batch, compute.containers | https://www.anchorterminal.com/tools/runpod.min.md |
| Verda | D | 53.3 | compute.gpu, compute.serverless, compute.endpoints, compute.batch, compute.containers | https://www.anchorterminal.com/tools/verda.min.md |
| Cerebrium | D | 52.5 | compute.gpu, compute.serverless, compute.endpoints, compute.batch, compute.containers | https://www.anchorterminal.com/tools/cerebrium.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)