The canonical rate table — every observed quote, its level, panel disposition and unchanged-days count, read against the assessed prints. This is the public panel the participation path contributes to; where the panel is thin, the prints above say so rather than carrying a number.
| Provider | Offer basis | $/GPU-hr | Availability | vs GAP | Observed | |
|---|---|---|---|---|---|---|
RunPod GAP spot panel register · posted, stock-gated | posted list · stock-gated | $2.69 | STOCK: LOWBENCHED | — | 2026-11-10 | runpod.io → |
ThunderCompute observed · outside benchmark panel | posted list | $2.95 | NO SIGNALUNREGISTERED | — | 2026-11-10 | thundercompute.com → |
Hyperstack GAP spot panel register | public on-demand list | $3.20 | ON-DEMAND; PUBLIC LIST; QUALIFYING | — | 2026-11-10 | hyperstack.cloud → |
Lambda GAP reserved-1Y panel · spot leg registered | 1-yr committed / on-demand list | $4.69 | ON-DEMAND INSTANCE; PUBLQUALIFYING | — | 2026-11-10 | lambda.ai → |
Lambda GAP reserved-1Y panel · spot leg registered | 1-yr committed / on-demand list | $4.79 | ON-DEMAND INSTANCE; PUBLQUALIFYING | — | 2026-11-10 | lambda.ai → |
Lambda GAP reserved-1Y panel · spot leg registered | 1-yr committed / on-demand list | $4.89 | ON-DEMAND INSTANCE; PUBLQUALIFYING | — | 2026-11-10 | lambda.ai → |
| Provider | Model | $/GPU-hr | Availability | Observed | |
|---|---|---|---|---|---|
RunPod GAP spot panel register · posted, stock-gated | H100 PCIe | $1.99 | STOCK: LOW | 2026-11-09 † | runpod.io → |
Hyperstack GAP spot panel register | H100 80GB | $2.50 | ON-DEMAND; PUBLIC LIST; | 2026-11-10 | hyperstack.cloud → |
Vast.ai GAP spot panel · executable-ask leg | H100 PCIE 80GB | $2.54 | BOOKABLE · 1× | 2026-11-08 † | vast.ai → |
RunPod GAP spot panel register · posted, stock-gated | H100 NVL | $2.59 | STOCK: LOW | 2026-11-09 † | runpod.io → |
Hyperstack GAP spot panel register | H100 NVLink 80GB | $2.60 | ON-DEMAND; PUBLIC LIST; | 2026-11-10 | hyperstack.cloud → |
Vast.ai GAP spot panel · executable-ask leg | H100 NVL 94GB | $3.39 | BOOKABLE · 1× | 2026-11-10 | vast.ai → |
| Provider | Commitment basis | $/GPU-hr | vs R1Y final print 2026-08-21 | Observed | |
|---|---|---|---|---|---|
Lambda GAP reserved-1Y panel · spot leg registered | 1-click-cluster committed (quote as of 07-24) | $5.85 | −$0.61 | 2026-11-10 | lambda.ai → |
Azure GAP reserved-1Y panel | ND96is_H100_v5 · eastus · 1-yr term | $7.08 | +$0.61 | 2026-11-10 | azure.microsoft.com → |
| Provider | Model | $/GPU-hr | Availability | Observed | |
|---|---|---|---|---|---|
RunPod GAP spot panel register · posted, stock-gated | H200 SXM | $3.59 | STOCK: LOW | 2026-11-10 | runpod.io → |
Vast.ai GAP spot panel · executable-ask leg | H200 NVL 141GB | $3.95 | BOOKABLE · 2× | 2026-11-10 | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | H200 NVL 141GB | $4.00 | BOOKABLE · 1× | 2026-11-10 | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | H200 NVL 141GB | $4.39 | BOOKABLE · 4× | 2026-11-06 † | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | H200 141GB | $5.26 | BOOKABLE · 2× | 2026-11-06 † | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | H200 141GB | $5.72 | BOOKABLE · 1× | 2026-11-10 | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | H200 141GB | $5.86 | BOOKABLE · 4× | 2026-11-04 † | vast.ai → |
RunPod GAP spot panel register · posted, stock-gated | B200 | $5.98 | STOCK: LOW | 2026-11-10 | runpod.io → |
Vast.ai GAP spot panel · executable-ask leg | B200 180GB | $7.50 | BOOKABLE · 2× | 2026-11-05 † | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | B200 180GB | $7.50 | BOOKABLE · 1× | 2026-11-10 | vast.ai → |
Vast.ai GAP spot panel · executable-ask leg | B200 180GB | $8.50 | BOOKABLE · 8× | 2026-11-04 † | vast.ai → |
Today, The Grid earns nothing from any provider on this page. No referral agreements exist; every link above is a plain link. If referral links are ever added, this band will say so by name, state what The Grid earns, and the following rules apply from the first dollar:
No payment influences assessment, panel inclusion, or coverage. Referral fees are permitted solely on this separated comparison surface under public disclosure and identical-terms rules. Assessment code cannot read referral data — the separation is structural. If referral income ever conflicts with assessment honesty, the referral surface is killed first; the print survives.
Identical terms are offered to every listed provider. Ordering is by observed price within class — there is no field in this system that can express paid placement.