Home · Reports · Vendor sample reports · Lium
Lium GPU price sample report
A one-page sample from Lantern Works, generated from the public price feed and daily snapshots. Rankings are among tracked vendors only, not the whole market. Price data is licensed CC BY 4.0 (attribution: Lantern Works).
Vendor pricing page used as the source for current rows: https://lium.io/.
Rank per GPU model
Rank is among tracked vendors with a comparable rate for the same published GPU name: a single-GPU, headline-eligible, non-reserved, non-serverless, non-interruptible offer. Lower hourly equivalent wins. Affiliate status is not a ranking factor. Multi-GPU, serverless, reserved, or interruptible offers are listed but not given a rank in this table, with the reason shown. If the feed has no hourly figure, it is listed as no figure recorded; none is invented.
| GPU | This vendor | Rank | Cheapest tracked | Read | Evidence |
|---|---|---|---|---|---|
| H200 NVL | $3.75/hr · Marketplace secure node · NVL | 1 of 1 | $3.75 · Lium |
| |
| RTX 3090 | $0.16/hr · Marketplace secure node | 1 of 2 | $0.16 · Lium |
| |
| B200 | $5/hr · Marketplace spot node (interruptible) | 2 of 4 | $5.98 · RunPod |
| |
| H100 PCIe | $1.994/hr · Marketplace secure node · PCIe | 2 of 5 | $1.99 · RunPod |
| |
| RTX 4090 | $0.38/hr · Marketplace secure node | 2 of 7 | $0.34 · RunPod |
| |
| RTX PRO 6000 Server | $1.7/hr · Marketplace spot node (interruptible) | 2 of 2 | $1.69 · RunPod |
| |
| H100 SXM | $2/hr · Marketplace spot node (interruptible) · SXM | not ranked (interruptible/spot) | — |
| |
| H200 | $3.79/hr · Marketplace secure node | not ranked (sold only as a 8-GPU minimum) | — |
| |
| L40 | $0.33/hr · Marketplace secure node | not ranked (sold only as a 4-GPU minimum) | — |
| |
| L40S | $0.38/hr · Marketplace secure node | not ranked (sold only as a 4-GPU minimum) | — |
| |
| RTX 5090 | $0.4/hr · Marketplace spot node (interruptible) | not ranked (interruptible/spot) | — |
| |
| RTX 6000 Ada | $0.79/hr · Marketplace secure node | not ranked (sold only as a 8-GPU minimum) | — |
| |
| RTX A6000 | $0.4/hr · Marketplace secure node | not ranked (sold only as a 4-GPU minimum) | — |
|
Undercut in the last 14 days
History is incomplete for this window. No daily snapshot was published for 2026-11-07, 2026-11-08. Undercut counts use only the 12 snapshot days that exist (2026-11-06 … 2026-11-19). No snapshot exists for those days, so undercut counts do not include them and do not assume prices held steady. No figures were invented for those days.
An undercut observation is one snapshot day and GPU name where this vendor had a comparable ranked 1-GPU, non-spot, non-reserved rate and at least one other tracked vendor was strictly cheaper the same day. Counts are observations, not distinct hours. Days with no comparable ranked rate for this vendor are not counted as undercuts.
48 undercut observations across 12 snapshot days where this vendor had a comparable ranked rate on A100 SXM, B200, B300, H100 PCIe, H100 SXM, H200, H200 NVL, L40S, RTX 3090, RTX 4090, RTX 5090, RTX 6000 Ada, RTX A6000, RTX PRO 6000 Server.
| Undercut by | Observations | GPU names |
|---|---|---|
| RunPod | 47 | B200, B300, H100 PCIe, H200, RTX 4090, RTX 5090, RTX 6000 Ada, RTX A6000, RTX PRO 6000 Server |
| CloudRift | 3 | L40S, RTX 5090 |
| Thunder Compute | 3 | RTX A6000 |
| Lambda | 3 | B200 |
| Clore.ai | 2 | RTX 5090 |
| Vast.ai | 1 | RTX PRO 6000 Server |
| Verda (formerly DataCrunch) | 1 | B300 |
Chart
Current hourly equivalent (amber) next to the cheapest tracked comparable rate (gray), for each GPU this vendor lists that is in the comparable set. The table above is the source; the chart does not add information.
Check against the vendor pricing page
Matches the vendor page. Fetched https://lium.io/ (HTTP 200); every current published hourly figure (H100 PCIe, H100 SXM, RTX PRO 6000 Server, B200, H200, RTX 4090, RTX 6000 Ada, RTX A6000, L40S, RTX 3090, RTX 5090) appears in the response body.
Disclosure: some vendor links on Lantern Works are referral/affiliate links (currently RunPod, Thunder Compute, Vast.ai, Novita AI, Clore.ai, Ucartz and 99Stack Cloud). Lantern Works may earn a commission. Affiliate status does not affect rankings.
Report a correction: report a price. See also the methodology and the gpu-prices.json feed.