L40S vs Tesla T4
L40S starts at $0.97/GPU-hour on-demand; Tesla T4 starts at $0.35/GPU-hour on-demand. Compare the exact variants, memory and published specifications below. Peak compute figures do not establish application speed.
L40S
Tesla T4
Spec sheet
| L40S | Tesla T4 | |
|---|---|---|
| Spec variant | L40S | Tesla T4 |
| Architecture | Ada Lovelace | Turing |
| VRAM (spec variant) | 48 GB | 16 GB |
| FP16 TFLOPS (dense) | 362 | 65 |
| Memory bandwidth | 864 GB/s | 320 GB/s |
| TDP | 350 W | 70 W |
| Interconnect | PCIe Gen4 | PCIe Gen3 |
| Launched | 2023 | 2018 |
Rental price per manufacturer-rated FP16 compute: L40S $2.68 per PFLOP-hour vs Tesla T4 $5.38. Formula: cheapest $/hr ÷ dense FP16 PFLOPS (1 PFLOP = 1000 TFLOPS). This uses L40S and Tesla T4, whose datasheets may differ from the variants above. It is not measured job throughput. Specs: L40S manufacturer, Tesla T4 manufacturer.
What they cost to rent (2026-10-02)
| L40S | Tesla T4 | |
|---|---|---|
| Cheapest on-demand $/hr | $0.97 | $0.35 |
| Cheapest spot/community $/hr | $0.79 | $0.199 |
| Cheapest source | Massed Compute | Google Cloud |
| Providers offering it | 16 | 5 |
Lowest recorded on-demand rate per configuration; highlighted = cheaper. Spot/marketplace tiers shown on the individual pages.
FAQ
Tesla T4 is cheaper: the L40S rents from $0.97 per GPU-hour and the Tesla T4 from $0.35 on-demand, updated 2026-10-02.
The largest recorded L40S variant has 48 GB and the largest Tesla T4 variant has 16 GB. Other variants may have less memory. The L40S has more.
Compare more rental options: the screener · price movers · methodology.