Modal Nvidia A100, 40 GB vs Modal Nvidia L40S.
machine vs machine · prices as of
Two machines from Modal's own catalogue, the nearest to each other on price. Modal Nvidia L40S takes 3 of 4 rounds against Modal Nvidia A100, 40 GB. It is also the cheaper of the two by the hour, at $1.95/GPU-hr against $2.10/GPU-hr.
Nvidia A100, 40 GB
$2.10/GPU-hr
$2.10/hr for the machine · $353 a week
1x A100 40GB SXM · 40 GB each · $0.0525/GB-hr · Serverless
Nvidia L40S
$1.95/GPU-hr
$1.95/hr for the machine · $328 a week
1x L40S · 48 GB each · $0.0407/GB-hr · Serverless
Price per GPU-hour (lower wins)
$2.10/GPU-hr
$1.95/GPU-hr
Interruptible rate (lower wins)
not published
not published
Price per GB of VRAM (lower wins)
$0.0525/GB-hr
$0.0407/GB-hr
VRAM per GPU
40 GB
48 GB
GPUs per instance
1x A100 40GB SXM
1x L40S
vCPU
not stated
not stated
RAM
not stated
not stated
Included storage
none · volumes extra
none · volumes extra
Regions listed
regions not stated
regions not stated
What a run costs on each
- An hour of the machine$2.10/hr vs $1.95/hr
- A day$50 vs $47
- A week$353 vs $328
- A week on spotnot published vs not published
Modal then Modal, at the published rate with no commitment discount
Go further
Add a third or fourth machine, or switch the weighting, in the tool.
Frequently asked
- Which is cheaper, Modal Nvidia A100, 40 GB or Modal Nvidia L40S?
- Modal Nvidia L40S, at $1.95/GPU-hr against $2.10/GPU-hr, both the on-demand rate for one GPU for one hour, read on 17 Sept 2026. On the whole machine that is $1.95/hr against $2.10/hr.
- Which holds more GPU memory?
- Modal Nvidia L40S, at 48 GB across the instance against 40 GB. Per board it is 40 GB against 48 GB.
- What would a week cost?
- $353 for Modal Nvidia A100, 40 GB and $328 for Modal Nvidia L40S, at 168 hours of the whole machine at its published rate, with no commitment discount. Neither publishes an interruptible rate.
- What does the hourly price leave out?
- Modal Nvidia A100, 40 GB: disk none, billed separately, no regions published, billed per second. Modal Nvidia L40S: disk none, billed separately, no regions published, billed per second. Egress and tax are extra on both.
Every rate is the on-demand price for one GPU for one hour, read 17 Sept 2026 from each provider's own page, so a node of eight is compared at what one of its GPUs costs. Storage, egress and tax are extra unless a row says otherwise, and reserved or committed rates are not indexed. Bars are scaled to the larger value in each row.