NVIDIA L4 vs NVIDIA L40, AI & Machine Learning Comparison
NVIDIA L40 wins 10 of 11 benchmarks, averaging 150.6% faster.
NVIDIA L40's numbers are anchored estimates calibrated against our measured cards, pending first-party measurement. Treat small gaps as ties.
What the numbers say
The gap is widest in Qwen3-32B, where NVIDIA L40 leads by 174% (12.42 vs 34.08 tok/s); the closest fight is Stable Diffusion XL (116% apart); VRAM decides part of this one: NVIDIA L40 runs 138 of our 12 AI workloads while the other card runs 11, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA L4 | NVIDIA L40 | Difference |
|---|---|---|---|
| Qwen3-4B tok/s | 84 | 211.33 | -60% |
| Llama 3.1 8B tok/s | 50.45 | 135.12 | -63% |
| Qwen2.5-Coder 14B tok/s | 27.46 | 74.27 | -63% |
| Qwen3-32B tok/s | 12.42 | 34.08 | -64% |
| Llama 3.3 70B tok/s | 0 | 0 | n/a |
| Stable Diffusion XL images/min | 5.18 | 11.2 | -54% |
| Z-Image Turbo images/min | 2.475 | 5.625 | -56% |
| FLUX.1 dev images/min | 0 | 2.464 | n/a |
| FLUX.1 Kontext dev images/min | 0 | 1.114 | n/a |
| Qwen-Image-Edit images/min | 0 | 0.48 | n/a |
| Wan 2.2 5B (720p) frames/s | 0.15 | 0.37 | -59% |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA L4 is faster on 0 of 3; NVIDIA L40 on 3.
| Workflow | NVIDIA L4 | NVIDIA L40 | Difference | Cost per run |
|---|---|---|---|---|
| 24-frame storyboard | 12.2 min | 7 min | NVIDIA L40 1.74x faster | $0.089 vs $0.080 |
| Full codebase review | 36.4 min | 13.5 min | NVIDIA L40 2.70x faster | $0.267 vs $0.155 |
| 10 short social clips | 60.3 min | 28.2 min | NVIDIA L40 2.14x faster | $0.442 vs $0.324 |
Renting by the hour, NVIDIA L40 finishes 3 of 3 cheaper. The quicker card is not automatically the cheaper way to get the work done.
Cost to rent
| Card | Per hour |
|---|---|
| NVIDIA L4 | $0.440 |
| NVIDIA L40 | $0.690 |
NVIDIA L4 is 1.57x cheaper per hour. Cheapest on-demand rate we see across RunPod and Vast.
Specifications compared
| NVIDIA L4 | NVIDIA L40 | |
|---|---|---|
| VRAM | 24GB | 48GB |
| Architecture | Ada Lovelace | Ada Lovelace |
| Memory bandwidth | 300 GB/s | 864 GB/s |
| Boost clock | 2,040 MHz | 2,490 MHz |
| TDP | 72 W | 300 W |
| Launch MSRP | $2,500 | $7,000 |
| Release | 2023-03-21 | 2022-10-13 |
FAQ
- Which is better for ai & machine learning: NVIDIA L4 or NVIDIA L40?
- NVIDIA L40 performs better for ai & machine learning, winning 10 of 11 benchmarks in our suite with an average 150.6% advantage.
- What are the main hardware differences between NVIDIA L4 and NVIDIA L40?
- NVIDIA L4 has 24GB VRAM and a 72W TDP, while NVIDIA L40 has 48GB VRAM and a 300W TDP.
- Does VRAM matter more than speed between NVIDIA L4 and NVIDIA L40?
- For AI, yes, NVIDIA L40 fits 24 more of our 12 workloads than NVIDIA L4. A model that exceeds VRAM doesn't run slower, it doesn't run at all, so the card that fits the model wins that workload outright.
- Where is the biggest performance difference between NVIDIA L4 and NVIDIA L40?
- Qwen3-32B: NVIDIA L40 leads by roughly 174% (12.42 vs 34.08 tok/s) in our testing.
NVIDIA L4 full review · NVIDIA L40 full review · All AI & Machine Learning rankings