Question 1

What AI models can run on L4?

Accepted Answer

The L4 (24 GB VRAM) can run models including Llama-2-7b-chat, Llama-3.2-1B-Instruct-SpinQuant_INT4_EO8, FastContext-1.0-4B-RL, Llama-3.2-1B-Instruct-QLORA_INT4_EO8, HARC-Llama-3.1-8B-Instruct and more — 10 compatible models indexed so far. Compatible models are determined by whether the model's memory footprint fits within the GPU's 24 GB VRAM, potentially using quantization (Q8, Q6, or Q4) to reduce memory usage.

Question 2

What is the cheapest cloud option for L4?

Accepted Answer

The cheapest on-demand option for the L4 is RunPod at $0.39/hr (1x_L4_SECURE). Spot/preemptible pricing is available at $0.39/hr. We track pricing across 128 providers to help you find the best deal.

Question 3

Is L4 good for fine-tuning?

Accepted Answer

The L4 can handle fine-tuning with some constraints. With 24 GB of VRAM, it's suitable for LoRA/QLoRA fine-tuning of models up to ~7B parameters. Full fine-tuning is limited to smaller models (~1-3B parameters). Techniques like gradient checkpointing and mixed precision help maximize what fits.

Question 4

How much VRAM does L4 have?

Accepted Answer

The L4 has 24 GB of VRAM (GDDR6). It also offers 300 GB/s memory bandwidth, 121 TFLOPS of FP16 compute, FP8 support for efficient inference.

Question 5

What quantization formats work on L4?

Accepted Answer

The L4 supports the following quantization formats: FP16 (full precision), FP8, Q8 (8-bit), Q6 (6-bit), Q4 (4-bit). FP16 works for models up to ~7B parameters. FP8 halves memory vs FP16 with minimal quality loss. Q4 quantization lets models up to ~81B parameters (by VRAM estimate) fit on the L4.

Provider	On-demand	Spot	Actions
RunPod	$0.390/hr	$0.390/hr	Rent Get $5–$500 free
Google Cloud	$0.558/hr	$0.195/hr	—
Amazon Web Services	$0.805/hr	$0.234/hr	—

L4

Specifications

Cloud pricing

Compatible models

Compare this GPU

Frequently asked questions

VRAM	24 GB
Memory type	GDDR6
Bandwidth	300 GB/s
FP16 TFLOPS	121
FP8 support	Yes

L4