L4 vs L40
L4 starts at $0.44/GPU-hour on-demand; L40 starts at $0.79/GPU-hour on-demand. Compare the exact variants, memory and published specifications below. Peak compute figures do not establish application speed.
Spec sheet
| L4 | L40 | |
|---|---|---|
| Spec variant | L4 | L40 |
| Architecture | Ada Lovelace | Ada Lovelace |
| VRAM (spec variant) | 24 GB | 48 GB |
| FP16 TFLOPS (dense) | 121 | 181 |
| Memory bandwidth | 300 GB/s | 864 GB/s |
| TDP | 72 W | 300 W |
| Interconnect | PCIe Gen4 | PCIe Gen4 |
| Launched | 2023 | 2022 |
Rental price per manufacturer-rated FP16 compute: L4 $3.64 per PFLOP-hour vs L40 $4.36. Formula: cheapest $/hr ÷ dense FP16 PFLOPS (1 PFLOP = 1000 TFLOPS). This uses L4 and L40, whose datasheets may differ from the variants above. It is not measured job throughput. Specs: L4 manufacturer, L40 manufacturer.
What they cost to rent (2026-10-02)
| L4 | L40 | |
|---|---|---|
| Cheapest on-demand $/hr | $0.44 | $0.79 |
| Cheapest spot/community $/hr | $0.105 | $0.69 |
| Cheapest source | Jarvislabs | Thunder Compute |
| Providers offering it | 10 | 6 |
Lowest recorded on-demand rate per configuration; highlighted = cheaper. Spot/marketplace tiers shown on the individual pages.
FAQ
L4 is cheaper: the L4 rents from $0.44 per GPU-hour and the L40 from $0.79 on-demand, updated 2026-10-02.
The largest recorded L4 variant has 24 GB and the largest L40 variant has 48 GB. Other variants may have less memory. The L40 has more.
Compare more rental options: the screener · price movers · methodology.