L4
See rental prices ↓Current median: $0.79/hr across 14 providers (observed), 0.00 (0.0%) vs yesterday. Typical range: $0.29–$0.95/hr (P10–P90). 30-day trend unlocks after 30 days of captures
TDP 72W · 60.6 TFLOPS · 24GB VRAM · 300 GB/s bandwidth
The L4 has a 72W TDP, 60.6 TFLOPS of FP16 compute, 24GB of GDDR6 VRAM, 300 GB/s memory bandwidth.
Extremely low TDP (72W). Ideal for high-density inference racks. Best performance-per-watt in its class.
L4 currently rents for a median of $0.79/hr across 14 providers (observed on-demand prices). L4 has 24GB of memory. Compare its specs and listed cloud rental prices below. The workload must fit in VRAM with room for runtime memory; a listed price does not confirm stock.
L4 Overview
The NVIDIA L4 is an Ada Lovelace data-center GPU optimized for efficiency rather than maximum model size. It provides 24GB of GDDR6 and 60.6 TFLOPS of FP16/BF16 performance in an exceptionally low 72W envelope, which makes dense, power-conscious inference deployments practical.
The 300 GB/s memory bandwidth is modest, so autoregressive models that repeatedly stream large weight sets will become memory-bound sooner than on HBM GPUs. Its 24GB capacity is a useful fit for compact models and quantized serving, but it does not leave much headroom for large FP16 weights, long contexts, or oversized batches.
The L4 excels at efficient inference, edge-adjacent deployment, video and vision pipelines, and high-density serving of modest models. It is not designed for large-model training, and its lack of NVLink means each card remains an independent 24GB resource.
Memory
Compute Performance
Hardware Specifications
Chip
Processors
Clocks
Memory
Power
Relative Performance
Relative to highest-spec GPU in database
Limitations
Live Cloud PricingOn-demand hourly rates
| Provider | On-Demand | Spot | Region | Availability |
|---|---|---|---|---|
BESTReplicate | $0.23/hr | — | US | unknown |
| $0.55/hr | — | APAC/ME/LATAM | unknown | |
| $0.80/hr | — | US/EU | unknown | |
| $0.79/hr | — | US | unknown | |
| $0.80/hr | — | US | unknown | |
| $0.30/hr | — | us-central1 | unknown | |
| $0.44/hr | — | US | unknown | |
| $1.24/hr | — | Paris, France, Warsaw, Poland | unknown | |
| $0.80/hr | — | us-east | unknown | |
| $0.89/hr | — | US/EU | unknown | |
| $0.32/hr | — | EU-West | unknown | |
| $0.95/hr | — | Various | unknown | |
| $0.59/hr | — | US | unknown | |
| $0.95/hr | — | EU/US | unknown | |
| $0.51/hr | — | us-east-1 | unknown | |
| $0.09/hr | — | Japan, JP | unknown | |
| $0.28/hr | — | EU-West | unknown |
Job Cost Estimator
At median $0.79/hr across 14 providers (observed on-demand). Spot pricing typically 40–70% lower.
Estimates only. Actual cost depends on provider, region, availability, and utilization. Use the GPU cost calculator for custom inputs.
Compare L4 vs…
Use Case Guidance
LLM Model Size Guidance
Estimates only. Actual capacity depends on context length, KV cache, and framework overhead.
Related Guides
LLM APIs Running on This GPU Class
Providers that serve frontier LLM inference on Ada Lovelace-class hardware.
L4 vs Alternatives — Spec Comparison
| Spec | L4 this | RTX 3080 | A16 | RTX 4070 |
|---|---|---|---|---|
| VRAM | 24GB GDDR6 | 10GB GDDR6X | 64GB GDDR6▲ | 12GB GDDR6X |
| Memory Bandwidth | 300 GB/s | 760 GB/s▲ | 432 GB/s | 504 GB/s |
| FP16 TFLOPS | 60.6 | 59.6 | 62.4▲ | 58.2 |
| BF16 TFLOPS | 60.6 | 59.6 | 62.4▲ | 58.2 |
| FP8 TFLOPS | — | — | — | — |
| INT8 TOPS | 242 | 119 | 250▲ | 116 |
| TDP | 72W | 320W▲ | 250W | 200W |
| Process Node | TSMC 4N | Samsung 8nm▲ | Samsung 8nm▲ | TSMC 4N |
| Architecture | AD104▲ | GA102 | GA102 | AD104▲ |
| Release Year | 2023▲ | 2020 | 2021 | 2023▲ |
| Max model (FP16) | ~12B params | ~5B params | ~32B params▲ | ~6B params |
| Max model (INT4) | ~48B params | ~20B params | ~128B params▲ | ~24B params |
Compare L4
Related GPUs
L4 Frequently Asked Questions
What is the L4 TDP?
The L4 has a TDP (Thermal Design Power) of 72W.
How many TFLOPS does the L4 have?
The L4 delivers 60.6 TFLOPS of FP16 compute and 60.6 TFLOPS BF16.
How much VRAM does the L4 have?
The L4 has 24GB of GDDR6 VRAM.
What is the L4's memory bandwidth?
The L4 has a memory bandwidth of 300 GB/s.
L4 User Reviews
L4 User Reviews
Ready to rent?
Compare L4 prices across 130+ providers
Live on-demand & spot rates · monthly cost estimates · availability status
LLM Self-Hosting Guide
Best LLM models to run on the L4
Low-power GPU for efficient 7B model serving at scale