Cloud GPU cluster
CoreWeave
GPU cloud with three inference paths: curated serverless token APIs, managed dedicated custom-weight inference and self-managed serving on CoreWeave Kubernetes Service. Cluster capacity also has on-demand, spot and reserved options.
Product facts checked October 2, 2026Editorial scores last reviewed September 3, 2026
Some outbound links use a first-party redirect hop for click counting. Commission is only claimed when a partner programme is active for that specific link — most vendor hops here are not paid placements. Affiliate disclosure.
Marketplace referrals (for example RunPod or Vast) may return credits or kickbacks when a programme is active — that is a material connection even when it is not a cash CPA.
Best for
Teams choosing between managed custom-weight inference, serverless model APIs and full GPU-cluster control
Watch out
Check the live SKU, contract and billing unit. CoreWeave reports a 10x tokens/sec-per-megawatt result for Vera Rubin NVL72 versus GB200 NVL72 on DeepSeek R1 at matched per-user interactivity; the result is workload-specific. Its October 2 price table lists 8-GPU nodes at $49.24/hr (H100), $50.44 (H200) and $68.80 (B200); inference-single rates are separately billed per GPU.
| Billing | Serverless per-token; Dedicated Inference per GPU-hour; CKS/cloud capacity uses on-demand, spot, reserved or Flex terms. |
|---|---|
| GPU choice | Dedicated Inference explicitly chooses GPU class and zone; serverless manages hardware; CKS exposes the full cluster serving stack. |
| Cold start | Dedicated model deployments load weights and need ready replicas; self-managed clusters require provisioning and your serving stack. No universal startup delay is established. |
| Model access | Serverless has a curated OSS/LoRA catalog. Dedicated Inference serves your weights from CoreWeave AI Object Storage through supported managed runtimes; CKS leaves model/runtime management to you. |
Sources for the dated product and announcement claims
- Current product modes, billing and deployment documentation — source checked October 2, 2026
- Current product modes, billing and deployment documentation — source checked October 2, 2026
- Current product modes, billing and deployment documentation — source checked October 2, 2026
- CoreWeave scoped DeepSeek R1 evaluation — source checked October 2, 2026
- CoreWeave current pricing units — source checked October 2, 2026
Fit detail
When CoreWeave is the right shortlist — and when it is not
Use these lists to kill bad comparisons early, before you compare logos.
Ideal for
- Managed custom/fine-tuned serving with explicit GPU and zone choices
- Serverless APIs when the supported catalog fits
- Large self-managed training/inference clusters with capacity contracts
Usually not ideal for
- Assuming every CoreWeave path requires Kubernetes operations
- Expecting arbitrary weights on the curated serverless API
- Deploying dedicated weights without enabled access, GPU quota and readable Object Storage artifacts
Spend & ops
How cost behaves — and what breaks first
No invented $/hour quotes. These notes explain the failure modes that show up on the first real invoice.
Cost mental model
Choose the product first: serverless token usage, managed dedicated GPU-hours or CKS capacity. Dedicated rates track chosen GPU/capacity and the product page says no extra service markup or ingress/egress fees; other storage/capacity contracts must be checked separately. Utilization and replica count drive dedicated spend.
Ops notes
- Dedicated Inference requires enabled account access, per-instance-type GPU quota and an Inference Admin API token.
- Upload weights to CoreWeave AI Object Storage; Dedicated Inference does not pull them directly from Hugging Face.
- Review bucket policy access for both the inference service and your own users; replacing implicit access with a service-only policy can lock organization users out.
Fit scores
CoreWeave on the decision axes
Same editorial scale as the hub chart — useful for shortlists, not as a price quote.
CoreWeave
- GPU choice8/10
- Time-to-serving (editorial)5/10
- Price clarity5/10
- Production ops8/10
- Open-model breadth7/10
Among all 16 providers
GPU choice
Pick exact GPUs
Time-to-serving (editorial)
Warm-path fit
Price clarity
Easy to forecast
Production ops
Less DIY ops
Open-model breadth
Catalog depth
Scores are editorial planning ratings (1–10) for product shape — not published $/hour quotes or vendor SLAs. Verify current pricing on each provider site.
Compare
CoreWeave vs alternatives
Open a head-to-head when you are deciding between product shapes.
Related
Keep the decision attached to the rest of the stack
Cloud rental is one path. Local VRAM and open-weight fit still matter.