AP
Agent Provider Depot
Opinionated picks for agents
OpenAPI
← All jobs

Multi-node GPU training

Train/fine-tune large models across many GPUs with fast interconnect.

GET /api/decide?jtbd=multi-node-trainingAPI embedCoreWeave vs LambdaCoreWeave vs AWS GPU (EC2/P5/SageMaker)

At a glance — Oct 2026

Updated 2026-10-27 · forward this page: /jtbds/multi-node-training#brief

Scores as of 2026-10-19 (8 days old).

Depot Pick
CoreWeave · 7.49/10 for this job (curated estimate, high confidence)

capability interconnect + reliability + maturity; pricePerf secondary.

Runner-up
Lambda · 7.87/10

Scores higher than the pick on: Price / performance, Developer & agent DX, Lock-in risk (higher = worse).

When not to use the pick
Single-node LoRA — RunPod/Modal. Already all-in on AWS — P5/SageMaker.
Lock-in / migration
Runner-up Lambda has lower stored lock-in risk (5.5 vs 6.5, 0–10, higher = more lock-in).
Sourced price (usd_per_gpu_hour)

Pick's lowest sourced SKU: NVIDIA HGX H100 inference single GPU $6.16 (usd_per_gpu_hour) — source, as of 2026-10-19

Cheapest matching SKU: Crusoe NVIDIA L40S 48GB $1.5 — source, as of 2026-10-19, re-checked 2026-10-27. Not the pick; not quality-adjusted.

Price history: first seen 2026-10-19 · last changed no change recorded · history JSON

Free option
No free resource lists this job. A cheap paid SKU is not a free option.
Failure modes to watch
  • Runaway spend: Set a budget and a tool-call cap per task. Alert when spend jumps. Stop the run at the cap instead of summarizing a partial trace as success.
Reported outcomes (submitted reports only)

This job: loading… JSON · reports never change scores · seed/test records excluded

Depot Pick (CoreWeave) on this job: loading… JSON · reports never change scores · seed/test records excluded

Same-job, same-unit comparisons: CoreWeave vs Lambda · CoreWeave vs Crusoe

Cite: Agent Provider Depot, Depot Pick for multi-node-training: coreweave, scores as of 2026-10-19. Agents: GET /api/decide?jtbd=multi-node-training

All sourced prices, cheapest first

Depot Pick

CoreWeave

capability interconnect + reliability + maturity; pricePerf secondary.

Pricing (directional): On-demand H100 ~$6+/hr class historically; B200 higher; Kubernetes-native (CKS); often sales for large capacity. Public company.

Cite: Depot Pick for multi-node-training: coreweave
Overall
7.5
curated_estimate
confidence: high · as of 2026-10-19
Capability fit
9.2
Price / performance
5.5
Latency / reliability
9.0
Developer & agent DX
7.5
Privacy / data retention
8.2
Lock-in risk (higher = worse)
6.5
Maturity / track record
8.8

Lock-in risk: higher = more lock-in (worse). Overall already inverts this dimension.

Sources: CoreWeave: https://www.coreweave.com

Enterprise training/inference fleet pick. Skip if you just need a single H100 for an afternoon — use RunPod.

Quote for Multi-node GPU training

Depot Pick stays CoreWeave. Prices below are sourced SKU rows only (as of 2026-10-27). Input and output are not blended. GET /api/quote?jtbd=multi-node-training

Pick is not a price winner. coreweave matching USD per GPU-hour is 6.16; cheapest among candidates is crusoe at 1.5.

  • NVIDIA GH200 (1 GPU instance) usd_per_gpu_hour $6.5 source as of 2026-10-19
  • NVIDIA HGX B200 inference single GPU usd_per_gpu_hour $8.6 source as of 2026-10-19
  • NVIDIA HGX H100 inference single GPU usd_per_gpu_hour $6.16 source as of 2026-10-19
  • NVIDIA HGX H200 inference single GPU usd_per_gpu_hour $6.31 source as of 2026-10-19

5 of 9 candidates have no sourced price row. 5 are excluded from the USD per GPU-hour comparison (gpu_hour). min sourced GPU-hour SKU among candidates; not coerced to $/1M; per-second GPU was not included; not quality-adjusted

Sourced USD per GPU-hour for Multi-node GPU training

Candidates for this job that already have a matching sourced price (gpu_hour), cheapest first. As of 2026-10-27. Each figure is one SKU in that unit, not a blend of input and output, not a cross-modal rate, and not a Depot score. min sourced GPU-hour SKU among candidates; not coerced to $/1M; per-second GPU was not included; not quality-adjusted The Depot Pick is not a price winner.

ProviderUSD per GPU-hourSKUSourceAs of
Crusoe$1.5usd_per_gpu_hourNVIDIA L40S 48GBpricing page2026-10-19
Nebius$1.8usd_per_gpu_hourNVIDIA RTX PRO 6000pricing page2026-10-19
Lambda$2.79usd_per_gpu_hourNVIDIA A100 SXM 80GBpricing page2026-10-19
CoreWeaveDepot Pick$6.16usd_per_gpu_hourNVIDIA HGX H100 inference single GPUpricing page2026-10-19

Runners-up

Lambda

Best boring GPU cloud for research labs who want clusters without CoreWeave enterprise theater.

7.9
gputrainingclustersresearch

AWS GPU (EC2/P5/SageMaker)

Use when AWS is destiny. For greenfield agent GPUs, neoclouds win on price and speed-to-pod.

7.4
gpuawsenterprisesagemaker

Crusoe

Solid neocloud when you want SLA + energy story. Not the first stop for hackathon GPUs.

7.7
gpuenterprisesustainable

When NOT to use the top pick

Single-node LoRA — RunPod/Modal. Already all-in on AWS — P5/SageMaker.

Full ranking (score table)

Ranked with this JTBD's weights. Scores are curated estimates unless marked otherwise. Agents: prefer /api/decide then this table for tie-breaks.

#ProviderOverallFit$/perfLatencyDXAs of
1Lambda7.98.86.88.88.52026-09-14
2Nebius7.78.67.08.47.62026-09-14
3Crusoe7.78.67.28.67.42026-09-14
4Voltage Park7.68.67.88.46.82026-09-14
5CoreWeaveDepot Pick7.59.25.59.07.52026-10-19
6AWS GPU (EC2/P5/SageMaker)7.49.44.89.57.02026-09-14
7Anyscale7.38.26.58.07.52026-09-14
8GCP GPU (Compute/GKE)7.39.05.09.07.22026-09-14
9Azure GPU (N-series/ND)7.28.85.09.06.82026-09-14

Score methodology for this job

capability interconnect + reliability + maturity; pricePerf secondary.

  • capabilityFit: 26%
  • pricePerformance: 22%
  • latencyReliability: 10%
  • developerDx: 18%
  • privacyPosture: 8%
  • lockInRisk: 8%
  • maturity: 8%