GPU Reservation Desk
Sources & method

Follow the rate back to its source.

Every offer is a supplier’s statement. Source links show what we read, when we read it and the exact product scope.

Compare a quote, not just a number.

Keep GPU form factor, memory, count, region, billing interval, storage state and commitment together. A per-GPU rate is not a per-node total. No throughput, reliability or cheapest-provider ranking is claimed.

GB and GiB remain in the source’s units. Taxes and add-ons stay beside rates when known. “Unknown here” means this capture does not answer the question.

What buyers asked

Saved discussions raised capacity delays, stopped-state bills, cold starts, regional storage and VM image control. Those concerns shaped Capacity and Whole bill. A complaint is an attributed experience, not proof of today’s product behavior.

The offline collection contains 492 deduplicated text units from 164 indexed Reddit threads and no Firecrawl pages. The rule-based triage is search-selected and includes partial captures; it is not a survey or a count of buyers. We mapped seven reviewed examples to the new fields and claim no fully resolved product complaint.

Custom ISO and persistent boot controls on Vast, regional NFS compatibility on Lambda, and actual cold-start behavior remain unanswered.

What this comparison cannot tell you

Dates and history

Read dates below are actual source observations. Lambda’s announced effective date stays beside its rate. Search-cache conflicts are retained separately from observed offering changes. No recurring update job is installed.

Download claims, sources and history (JSON) ↓

RunPod · runpod-pricing

Vendor statement · Read 2026-11-13T03:49:54.556Z
https://www.runpod.io/pricing
Snapshot SHA256 0b8b9f8275f1dd21e5ada7d79e7ecc55f871cd9513351628b0bee203cd08f810 · Saved source text

GP-001 · Dedicated Pod — $2.89 / hour

H100 PCIe · 80 GB VRAM · 188 GB RAM · 16 vCPU · Secure Cloud. Per-second billing; no commitment. GPU price only; tax and storage total unknown.

Locator: text:998-1049; FAQ H100 PCIe
Exact claim SHA256: 0dab3bf50816bda839486f4c5bf89afeb04715dfa39e5ac0d5598392d524bd8a

GP-002 · Volume disk — $0.10 running / $0.20 idle per GB/month

Pod Volume Disk. Storage tier matters. Standard network storage has different rates.

Locator: Storage > Volume Disk
Exact claim SHA256: 2874376da68e446fd63cd8ea05c3231ee129b09d86f8d58838556a2640f32b81

GP-003 · Instant Clusters — Up to 64 GPUs; no commitment

Instant Clusters. Catalog statement, not an observation of available stock.

Locator: Clusters paragraph
Exact claim SHA256: 3813bf5e44ffed1c7911d4593cb52ea96fafeee7d865f60addbb3e6b021843c6

GP-004 · Reserved Clusters — Sales quote; SLA-backed uptime claimed

Reserved Clusters. No contract uptime target, remedy or regional stock captured.

Locator: Reserved Clusters
Exact claim SHA256: a94c24d83d2fe581fb76f4ecf8f9867c27d4bbedda5b1a2f4b343d41659c5dd9

CoreWeave · coreweave-pricing

Vendor statement · Read 2026-11-13T03:50:07.228Z
https://www.coreweave.com/pricing
Snapshot SHA256 e91391f2041ff3bfea341e16ad065491ddfc5da886d046c120c0d73184ca4c6b · Saved source text

GP-020 · On-demand node — $49.24 / node-hour

8 × HGX H100 · 80 GB per GPU · 128 vCPU · 2,048 GB RAM · 61.44 TB local storage. North America node price. The page’s $6.16 single-GPU inference price is restricted to inference-platform customers.

Locator: North America > NVIDIA HGX H100
Exact claim SHA256: 053e9dc36a9ede556fca51f8d0286e99d3a1d574b4fcdd7bfe51369df53ccb5e

GP-021 · Burst capacity — On-demand and Spot

On-demand GPU offering. Does not establish capacity or a reservation guarantee.

Locator: Opening > On-demand GPU instances
Exact claim SHA256: dd21453ba3a34cfa165125ba3d2f31d7a1762a62aafac124c3fa26256dcf961a

Hyperstack · hyperstack-pricing

Vendor statement · Read 2026-11-14T12:57:05.465Z
https://www.hyperstack.cloud/gpu-pricing
Snapshot SHA256 a19db5fdac3c4d07087a5229719760377a6737c3bdf211ad0b4b32a48bad25e6 · Saved source text

GP-005 · On-demand GPU — $3.20 / GPU-hour

H100 SXM · 80 GB VRAM · up to 24 pCPU and 240 GB RAM per GPU. Billed per minute. Maximum CPU/RAM, not a guaranteed chosen flavor.

Locator: L29,L54-L62
Exact claim SHA256: b2cc2c2dd7fc57e9432a59cee40852eeecf90e917a0bfcc789f198898a64d210

GP-006 · Ingress / egress — Free

Pricing page other-services traffic. Read alongside selected service and contract; storage and public IP separate.

Locator: L367-L379
Exact claim SHA256: 442e6667bf06e92c74c4609c4eccbd0e39c1f1fd67f65491d55610316f1c47fe

GP-007 · Shared storage volume — $0.000096774 / GB-hour

SSVs. Snapshot estimate varies with VM and attached volumes.

Locator: L347-L355
Exact claim SHA256: a4551947a4f118bdb007a1b64c4848c554751fd6399b45df2c5230a8dc59c33b

Nebius · nebius-pricing

Vendor statement · Read 2026-11-14T12:57:05.465Z
https://nebius.com/prices
Snapshot SHA256 a2f98969c6aaab753f48320d6479145e45775ebac0d9a2c0580931e75947cd46 · Saved source text

GP-008 · On-demand GPU — $4.50 / GPU-hour

HGX H100 · 16 vCPU · 200 GB RAM in the listing. Region-specific live stock and total quote unknown.

Locator: L107-L115
Exact claim SHA256: c80a589dbe23cb55fc0b695ea4a46531224f1ce541371f153c02f52be71f1182

GP-009 · Spot pricing — May change as often as every 15 minutes

Preemptible GPU spot prices, except flat-rate L40S. Published floors are not quotes or upper limits.

Locator: L146-L147
Exact claim SHA256: 9ee47b4016478b5d3954a1320fb6da4d633534d2458dc14a48700bcf4b5e2512

GP-010 · Shared filesystem — $0.0800 / GiB-month

Shared Filesystem by Nebius. WEKA, object and block storage have separate schedules.

Locator: L181-L185
Exact claim SHA256: b1567897a7f5e239bd0474dc3db0b62e5a649ec3e79127bda488bd6a1f043aa3

GP-011 · Object Standard egress — $0.0150 / GiB

Object storage Standard egress. Networking ingress/egress is separately listed free. Do not apply that to this storage tier.

Locator: L193-L203
Exact claim SHA256: c5b2ca3cc76d918cbbf67c5c8bf1ac9a3786845febc37c584692c669237462e6

Vast.ai · vast-pricing

Vendor statement · Read 2026-11-14T12:57:12.651Z
https://docs.vast.ai/guides/instances/pricing
Snapshot SHA256 101223e4bd78a26155c693e4829555bd94bb2c032e73ca99231c735279ea5ccc · Saved source text

GP-012 · Marketplace rental — Host-set rates

Vast marketplace. No exact live offer captured. Check GPU, host, location and both bandwidth rates.

Locator: L103-L117
Exact claim SHA256: 4cd027999205fdfee45ad2c5e94e5bafb9d3ca14020eafa3592fcd5235ad4fc5

GP-013 · Stopped instances — Storage keeps billing until deletion

Instances. Storage rates vary by host; stopping is not deletion.

Locator: L137-L141
Exact claim SHA256: 8d8b588d7fb39158ddccec089eb4b775afb56313aeca185f38c3e4721a4afc7d

GP-014 · Bandwidth — Upload and download charged per byte

Instance host rates. Check both rates in the selected offer.

Locator: L142-L144
Exact claim SHA256: 7cbb6b94becdded06b8513e8a0b528b5997e8159dd2b99e95790ec3219708b15

GP-015 · Interruptible rental — May be paused; low priority

Interruptible instances. A vendor price description, not a measured savings claim.

Locator: L145-L150
Exact claim SHA256: 0c611594cf643208f301c87fb8851e49c413e674ab91471a512c25d0be38aa7d

Crusoe · crusoe-platform

Vendor statement · Read 2026-11-14T12:57:12.651Z
https://www.crusoe.ai/cloud
Snapshot SHA256 b790ea5a84e1109818a65104d348f17db99b1e0a973da1480d3b69c348daf556 · Saved source text

GP-028 · GPU capacity — Spot, on-demand and reserved

Cloud GPU instances. Flattened pricing table cannot safely pair GPU rows/rates in capture. Exact GPU price left unknown.

Locator: L179-L184
Exact claim SHA256: e54d1379ce9faa19262ab17c96773cf17e4b83b264931c1cfcedd41cd8a3a0f7

GP-029 · Cluster networking — InfiniBand and Ethernet; zero egress fees claimed

Cloud networking statement. Not a network benchmark or an SLA.

Locator: L167-L172
Exact claim SHA256: b97ac345f161f00a42515b8a3bda2bb6f5b86a9ca5c4b8d9e75d267e48ece8e8

Crusoe · crusoe-pricing

Vendor statement · Read 2026-11-14T12:57:25.479Z
https://www.crusoe.ai/cloud/pricing
Snapshot SHA256 e5b4ed3f92bf9bd559d503abdaac394c18e00893f2ed9f032e5320846829da1e · Saved source text

GP-030 · Persistent / shared disk — $0.08 / $0.07 per GiB-month

Persistent disks and shared disks respectively. Separate from compute; no complete total or tax estimate.

Locator: L247-L263
Exact claim SHA256: c2631ed9066ed7756bc612c67db71ee9d6856a111276b72b5799e417bc2f212f

Together AI · together-clusters

Vendor statement · Read 2026-11-14T12:57:12.651Z
https://www.together.ai/gpu-clusters
Snapshot SHA256 a0eb950bec80c0e6eda645d153715effc254640c8e3958b0e6d5e524e6bb32df · Saved source text

GP-025 · On-demand cluster — $3.99 / GPU-hour

H100 · 80 GB per GPU · listed cluster scale starts at 8 GPUs. Do not treat as a single GPU rental. Live availability varies.

Locator: L247-L263
Exact claim SHA256: 5059ad89c06175b502334d5db7e38d8cdd5dce47ed02ac6d8f95af71698d629d

GP-026 · On-demand scale — Up to 256 GPUs; terminate anytime

On-demand clusters. Based on real-time availability.

Locator: L197-L219
Exact claim SHA256: c793a8e108cead4a3e85bbc08e490eb735327f30d0112a609fd6a2f991a38568

GP-027 · Reserved scale — Up to 4,000+ GPUs; upfront payment

Reserved clusters. Capacity locked for duration; quote must specify exact term.

Locator: L223-L240
Exact claim SHA256: dfbbc2958bd1e6df8c7f69a9f531f9b0add067ff0a329d64feb8cb6d84d04655

AWS · aws-p5

Vendor statement · Read 2026-11-14T12:57:25.479Z
https://aws.amazon.com/ec2/instance-types/p5/
Snapshot SHA256 c13575b2d283e224ea13a02d5e7f21af0d18704f4f5ed16d5c2f28968132bcf2 · Saved source text

GP-031 · P5 family — Up to 8 H100 GPUs · 640 GB total GPU memory

EC2 P5, not P5e/P5en H200 variants. Exact instance, region and price not captured.

Locator: L255-L258
Exact claim SHA256: 3ed8caf3238d19fe8e3bb6f7b0ad07cd22dcc8ac3cdce588c96c004a5f76b75c

GP-032 · P5 network — Up to 3,200 Gbps EFA

P5/P5e/P5en. Vendor maximum; not measured throughput or one GPU network guarantee.

Locator: L262-L267
Exact claim SHA256: e5909436601f8bc6fe748aecf893ba394137bf7ec648f86ffb4406e0b0abfbf5

Google Cloud · google-gpu

Vendor statement · Read 2026-11-14T12:57:41.899Z
https://docs.cloud.google.com/compute/docs/gpus
Snapshot SHA256 2de35a5df7571934cb3e3053eea2cdcc058aab8606f2244884dd24a18147dd68 · Saved source text

GP-033 · A3 High 8GPU — 8 × H100 · 640 GB total GPU memory

a3-highgpu-8g · 208 vCPU · 1,872 GB RAM · 6,000 GiB local SSD. Up to 1000 Gbps network. Region/price/stock unknown.

Locator: L691-L696
Exact claim SHA256: e2f922923b86501692f4ec92119c4b04b2ab8c3cc94998d3e34d57d66af7f2f6

GP-034 · A3 High 1/2/4GPU — Requires Spot or Flex-start

a3-highgpu-1g/2g/4g. Restriction is not stated here for 8 GPU variant.

Locator: L684-L690
Exact claim SHA256: aa563638115c960813ec3b9b5d21b585c84f16419ef74c5972f74a18d1a8e73e

Lambda · lambda-instances

Vendor statement · Read 2026-11-14T12:57:18.346Z
https://lambda.com/instances
Snapshot SHA256 1fb419c8fbd99ddc298f0fb28c8a583689d56ffc759c89c71de138cbff9e8126 · Saved source text

GP-023 · Instance access — Self-serve, first-come

1–8 GPU Instances. Minute billing and no egress fees claimed. Stock not observed.

Locator: L118-L146
Exact claim SHA256: 852eeba4c418fc5ebd043b67e939649e834e19c1ccc0db64d71a652916404b6d

Lambda · lambda-pricing-new

Vendor statement · Read 2026-11-14T12:57:25.479Z
https://lambda.com/pricing
Snapshot SHA256 d25a2d32113c2d94916d002d376c0a9e4bf18c28e31f4a8c5a4e28347506bc22 · Saved source text

GP-022 · Announced instance rate — $3.79 / GPU-hour from November 16

1 × H100 PCIe · 80 GB VRAM · 26 vCPU · 225 GiB RAM · 1 TiB SSD. Effective 2026-11-16 00:00 UTC plus applicable sales tax/VAT/GST. Current price requires account; do not use as today’s rate.

Locator: L181-L190 plus L166 effective-date note
Exact claim SHA256: 75d893248439d54660d50ee74aac274c646a3cbf62fa49ebb723e4244c8f4845

GP-024 · 1-Click H100 cluster — $6.16 / GPU-hour at 16 GPUs

16 × H100 · 2 weeks to 1 year. Plus applicable taxes. Different product from single GPU Instances; terms of one year or longer require a quote.

Locator: L142-L151
Exact claim SHA256: 2afa50011f09eec26217baaa636ba9dd90c74aad80c99a5d13a0928543ce1978

DigitalOcean / Paperspace · paperspace-pricing

Vendor statement · Read 2026-11-14T12:57:41.899Z
https://docs.digitalocean.com/products/paperspace/pricing/
Snapshot SHA256 b461d0f7a48aa9ae179aa8ce3846904c9281bc866d5a039f540caaf9750a8e1e · Saved source text

GP-035 · Legacy H100 Machine — $5.95 / machine-hour

Paperspace Machines in legacy console. GPU form factor and memory not specified on price page. Do not equate with GPU Droplets.

Locator: L48-L70
Exact claim SHA256: e9e18b14cfb074aef42d9ad750d1c3868d4f19ed63bdaea749929f3e57fe3903

GP-036 · When powered off — Compute stops; storage and IP still bill

Paperspace Machines. Hourly add-on charges continue until resources are destroyed. Non-GPU storage and public IP resources have maximum monthly charges.

Locator: L21-L31
Exact claim SHA256: 7d9355328fe4ad2f483446747883ef0974b0d49b73511396dd17a2e4e2788f36

GP-037 · Machine transfer — No ingress or egress charge

Paperspace Machines. Do not transfer this claim to DigitalOcean Droplets.

Locator: L26-L31
Exact claim SHA256: 6e862737556a33fddb54169d2bcb2582882e83a3b8836bff95b449fdbe464c11

DigitalOcean / Paperspace · paperspace-brand

Vendor statement · Read 2026-11-14T12:57:41.899Z
https://www.digitalocean.com/solutions/paperspace-joins-digitalocean
Snapshot SHA256 1aa67fb7d1f33f9c9d9083d1d16279653bf50fd31faad537f6965c513df0ee40 · Saved source text

GP-038 · Where to start — DigitalOcean directs new users to its AI/ML suite

DigitalOcean / legacy Paperspace. Existing users retain legacy products. No assertion of product retirement.

Locator: L100-L102,L224-L240
Exact claim SHA256: 6eb1fcb87c2733a827b089a87fb1b76e0080b8182c64679bfc59423dd80006fb

Earlier accepted facts

The five original RunPod records remain unchanged in the download under prior_records. Their earlier wording and candidate hashes are retained as history.

About & trust

GPU Reservation Desk is a local comparison prototype for ML infrastructure buyers. Vendor pages support documented offers; no independent speed or uptime tests were run. There are no paid placements or outbound lead forms in this build.

No audience, revenue or savings have been measured. This site does not collect submitted workloads or account credentials. Corrections can be made through the local project owner; no public support channel has been configured.