Rent an NVIDIA L4 for modern Ada-class inference with FP8 and AV1 NVENC. 24 GB Ada at 72 W passive — vLLM serving Llama-3 8B INT8 at ~950 tok/s with 16-request concurrency, three simultaneous 4K60 AV1 transcode streams, Florence-2 captioning at 28 images per second. Continuous-batching headroom prior-gen inference cards never had. Billed per-minute, paid in BTC, USDT/USDC or CLORE. The streaming and embedding card of choice for 2026 datacenter ops.
L4 is what happens when Nvidia takes 4090-class silicon, doubles the VRAM, adds ECC, locks the clocks for 24/7 operation, and ships it in a passive datacenter form factor.
24 GB Ada at 72 W passive — same form factor economics as a T4 but with FP8 tensor paths, AV1 NVENC, and enough VRAM to run vLLM with proper concurrency. The card datacenter operators standardized on for production inference and streaming transcode in 2026.
SDXL, Flux, Stable Video Diffusion, HunyuanVideo. RT cores 3rd-gen + OptiX make it the fastest non-HBM card for diffusion.
13B–34B QLoRA on a single card. 24 GB ECC means stable long runs without the corruption risk of consumer GDDR6X.
L4 delivers Hopper-class FP8 inference at a fraction of the H100 hourly rate. The right card when latency matters but you're not training from scratch.
// prices are spot-market lows · refreshed every 60 s
Every server is priced by its host. These are the live floors across the marketplace — you'll see hundreds of variants once you're in.
No sales call. No quota request. No three-week procurement. The first four commands are all you need.
Filter the marketplace by L4 48GB, country, GPU count, reliability score, network speed.
Choose a Docker image — PyTorch, vLLM, ComfyUI, Blender — or paste your own.
You get a public endpoint, an SSH key, and Jupyter on port 8888 in under 90 s.
Per-minute billing rounds to the second. Stop the instance and the meter stops with it.
Power-efficient (72 W vs 450 W), passively cooled, designed for 24/7 multi-tenant inference. Datacenter-validated for serving stacks like vLLM and Triton. The 4090 is faster per-card; the L4 is cheaper per-request at scale.
An L40S serving Llama-3 70B FP8 with vLLM and continuous batching pushes roughly 3,000-4,500 output tokens/second at batch saturation. At a $0.78/hr spot rate, that lands near $0.05-$0.07 per million output tokens before the 2.5% spot fee. PoH staking knocks the fee in half; reserved spot floors land you closer to $0.04/M. Numbers vary with prompt length and batch shape - benchmark on your traffic.
Yes. The inference tier (T4, L4, L40S, A10) is exactly what vLLM's PagedAttention and continuous batching are tuned for. L40S handles 70B FP8 single-card with KV-cache headroom; A10 and L4 serve 7B-13B at high throughput; T4 covers Whisper, embeddings, and 7B INT8. Pull the official vLLM Docker image, point it at your model, expose port 8000.
L40S has Ada FP8 tensor cores - the same architecture as H100 for inference math, at a fraction of the rental price. L4 also supports FP8. T4 and A10 are pre-FP8 but have INT8 (T4 added INT8 in Turing, A10 in Ampere) and excel at quantized 7B-13B serving. Pick L40S when FP8 throughput matters; pick A10 or T4 when $/request matters more.
On A10 or L4 with vLLM and batch-1, time-to-first-token for a 7B FP16 model lands around 80-150 ms; p99 inter-token latency is 25-40 ms. L40S with FP8 cuts both roughly in half. T4 doubles them. Real numbers depend on prompt length and concurrent batch size - low-batch interactive serving is fastest, high-batch saturation maximizes throughput.
MIG (Multi-Instance GPU) is supported on A100, A30, and H100/H200 - not on L4, L40S, T4, or A10. For consumer-tier multi-tenancy on the inference tier, run multiple model replicas inside a single Docker container or use container-level resource limits. If you need hardware-isolated MIG slices, rent A100 40GB and partition into up to 7 instances.
24 GB Ada at 72 W passive — the modern replacement for T4 with FP8, AV1 NVENC, and continuous-batching headroom.
24 GB Ada is the cheapest stable card for production 8B serving with batch >=16 concurrency.
Read the guide →L4's AV1 encoder and 72 W envelope make it the streaming transcode card of choice in 2026.
Read the guide →Vision-language captioning for stock-photo libraries — 24 GB fits Florence-2 large + 16 batch.
Read the guide →Side-by-side specs across the inference tier. Click any row to see that GPU.
Step-by-step guides verified on CLORE.AI hardware. Pick a workload, copy the docker image, ship in minutes.
Per-minute payouts in BTC, USDT, USDC or CLORE. No listing fee, no contracts, withdraw any time.
Hosts around the world are accepting workloads right now. Sign up, top up your wallet, and the next hour is yours.