Rent L4
Compare live on-demand and spot rental prices across 97+ cloud providers. Extremely low TDP (72W). Ideal for high-density inference racks. Best performance-per-watt in its class.
Choosing the right billing model for L4
Provision and terminate at any time. Ideal for development, short experiments, and workloads with unpredictable duration.
Instances can be reclaimed when demand spikes. Best for fault-tolerant batch jobs, training with checkpointing, and preprocessing.
Lock in a rate for 1–3 months. Right for sustained production inference or long training runs where cost predictability matters.
Best use cases
- Efficient inference
- Edge deployments
- Low-power serving
L4 — Specs & Benchmarks
Performance bars, compute tiers (FP32/FP16/BF16/FP8/INT8), memory specs, LLM model size guidance, and related GPU comparisons.
L4 Rental Guide
Rent an L4 when watts, rack density, and cost per inference matter more than training speed or large VRAM. It is particularly attractive for embeddings, vision, media pipelines, and smaller language models where its 24GB footprint has been validated in advance.
Spot L4 instances are a good match for asynchronous batch scoring, transcoding, and retryable inference queues. On-demand instances make more sense for lightweight, always-on APIs or edge-style services where capacity continuity is more valuable than an additional spot discount.
Do not compare the L4 only by FP16 TFLOPS with larger accelerators: its 300 GB/s bandwidth and 24GB capacity define its real LLM limits. Move to an A30 for more HBM bandwidth at modest power, or to L40S/A100-class hardware when model size, context length, or throughput expands.
Frequently Asked Questions
How much does it cost to rent a L4?
L4 on-demand rental prices vary by provider and region. On-demand rates typically range based on availability and provider margins — use the comparison table above to see current live rates across all providers. Spot instances are generally 40–70% cheaper than on-demand but can be interrupted. Monthly cost estimates (hourly rate × 730 hours) are shown in the table for sustained workloads.
Which cloud provider has the cheapest L4?
The cheapest L4 provider changes as providers update their pricing. The comparison table above shows live rates sorted by price, so the cheapest option is always at the top. Factors beyond headline price include region (latency to your users), availability (high/medium/low), and billing granularity (per-second vs per-hour minimums).
What can I run on a L4?
With 24GB of GDDR6, the L4 can run LLM models up to approximately 12B parameters at FP16, 24B at INT8, or 48B at INT4/GGUF quantization. Common workloads include: Efficient inference, Edge deployments, Low-power serving. Extremely low TDP (72W). Ideal for high-density inference racks. Best performance-per-watt in its class.
Should I use on-demand or spot pricing for L4?
Spot instances save 40–70% vs on-demand but can be interrupted when the provider needs capacity back. Use spot for: batch inference jobs, training runs with checkpointing, preprocessing pipelines, and any workload that can tolerate interruption and restart. Use on-demand for: production inference serving, interactive workloads, and jobs that cannot be interrupted. Most providers bill per second, so short on-demand jobs are not penalized by hourly minimums.
How does the L4 compare to the H100 for cloud rental?
The H100 80GB delivers 1,979 TFLOPS FP16 with 3,350 GB/s HBM3 bandwidth, compared to the L4's 60.6 TFLOPS FP16 and 300 GB/s bandwidth. The H100 is significantly more expensive — typically $2.50–$5.00/hr vs lower rates for the L4. For workloads that fit within 24GB and don't require FP8 precision, the L4 often delivers better cost-per-token than the H100.
What is the memory bandwidth of the L4 and why does it matter?
The L4 has 300 GB/s of memory bandwidth. For LLM inference, memory bandwidth is often more important than raw TFLOPS — each autoregressive token generation reads the full model weight matrix from VRAM, so bandwidth directly determines tokens-per-second throughput. Higher bandwidth means faster inference for the same model at the same batch size. For batch inference (processing many requests simultaneously), compute throughput becomes more important.
Can I use the L4 for Stable Diffusion or image generation?
Yes — the L4 is capable for Stable Diffusion and image generation workloads. Image generation is primarily FP32 and FP16 compute-bound, and the L4's 30.3 TFLOPS FP32 throughput determines images-per-second. The 24GB VRAM fits SDXL (requires ~6GB) and most ControlNet pipelines. For high-throughput image generation at scale, compare cost-per-image across providers using the GPU cost calculator.