NVIDIA / discrete

NVIDIA L4

24 GB decides what fits. 300 GB/s decides how fast it runs once it does.

Open in the calculator →Best models for 24 GB, on every card that size →

Quick answer

The NVIDIA L4 fits 214 of 320 sized models entirely in its 24 GB; the largest widely used one is Qwen3.6-35B-A3B (23.1 GB at Q4_K_M). Memory decides what fits; its 300 GB/s of bandwidth decides how fast it answers.

Largest popular model that fitsQwen3.6-35B-A3B36B params · Q4_K_M · needs 23.1 GB~110 tok/s Faster than you readOpen in the calculator →
Most downloaded that fitsQwen3.8-27B28B params · Q4_K_M · needs 18.6 GB~14 tok/s About reading paceOpen in the calculator →

At 8,192 tokens of context with the whole model in device memory. Speeds are estimates from memory bandwidth, not benchmarks run on this card.

NVIDIA L4 · 24 GB · 300 GB/s · Q4_K_M where published · 8,192 tokensSpecificationpublished dataDecode speedestimate

214 of 320 fit entirely

214 of the 320 models the engine can size fit entirely in device memory at Q4_K_M where it is published, otherwise the nearest published format, and 8,192 tokens. 19 run with some layers on system memory (32 GB assumed), and 87 do not run at all.

67%of the models the engine can size fit entirely

24GB of device memory

300GB/s memory bandwidth

19run with system memory

87do not run at 8,192 tokens

Featured models · Q4_K_M at 8,192 tokens, including offload

Model	Needs	Verdict	Decode	Calculator
Qwen3.5-2B2.3B parameters	2.32 GB	fits	~165 tok/s	Open →
NVIDIA-Nemotron-3-Nano-4B-BF164.0B parameters	3.46 GB	fits	~77 tok/s	Open →
Qwen3-VL-8B-Instruct8.8B parameters	7.04 GB	fits	~35 tok/s	Open →
Qwen3.5-9B9.7B parameters	7.05 GB	fits	~42 tok/s	Open →
gemma-4-12B-it12B parameters	9.54 GB	fits	~25 tok/sAbout reading pace	Open →
gpt-oss-20b21B parameters	13.8 GB	fits	~82 tok/s	Open →
Qwen3.8-27B28B parameters	18.6 GB	fits	~14 tok/sAbout reading pace	Open →
Qwen3.6-27B28B parameters	18.6 GB	fits	~14 tok/sAbout reading pace	Open →
Kimi-Linear-48B-A3B-Instruct49B parameters	30.4 GB	6 of 27 layers on system RAM (6.55 GB)	~73 tok/s with offload	Open →
Qwen3.8-Flash-Next180B parameters	111.6 GB	needs 89.8 GB of system RAM; 32 GB assumed	does not run	Open →
DeepSeek-V4-Flash-Vision-Exp305B parameters	188.1 GB	needs 165.5 GB of system RAM; 32 GB assumed	does not run	Open →
GLM-5.3-Flash321B parameters	198.3 GB	needs 175.3 GB of system RAM; 32 GB assumed	does not run	Open →

These examples are selected from prominent labs using the catalogue’s latest Hugging Face 30-day downloads and repository-creation freshness signal, with newer releases guaranteed a place. Offloaded rows assume 32 GB of system RAM, and a speed is only shown for a row that runs. Fitting in memory is not the same as loading: whether the runtime and version you have supports each architecture and format on this machine has not been tested here. Each “Open” link carries the same model, format, context and RAM into the calculator.

Fits entirely in NVIDIA L4 memory — 214 of 320 sized models

Every model that fitswhole model resident · offload off

The most demanding model that fits is Qwen2.5-VL-32B-Instruct at Q4_K_M: 23.5 GB of the 24 GB, leaving 0.52 GB spare.

Fully resident at 8,192 tokens (or the model’s own maximum, where that is shorter), offload off, at 300 GB/s. Each row is a run of the engine for this configuration; the rows start with current, prominent releases and “fits” is memory, not a tested runtime. Older or less prominent models remain available through this search and “Show all”.

Find a model in this list

Showing 40 of 214 models that fit.

Model	Format	Needs	Spare	Decode	Downloads / 30d	Calculator
Qwen3.8-27BQwen · 28B params	Q4_K_M	18.6 GB	5.45 GB	~14 tok/sAbout reading pace	6.9M	Open →
gemma-4-26B-A4B-itGoogle · 26B params	Q4_K_M	16.8 GB	7.21 GB	~77 tok/s	13M	Open →
gemma-4-31B-itGoogle · 31B params	Q4_K_M	22.2 GB	1.83 GB	~9.9 tok/sSlow	9.9M	Open →
Qwen3.5-9BQwen · 9.7B params	Q4_K_M	7.05 GB	17.0 GB	~42 tok/s	9M	Open →
Qwen3.5-4BQwen · 4.7B params	Q4_K_M	3.98 GB	20.0 GB	~76 tok/s	7.8M	Open →
Qwen3.6-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	0.90 GB	~110 tok/s	3.3M	Open →
gemma-4-12B-itGoogle · 12B params	Q4_K_M	9.54 GB	14.5 GB	~25 tok/sAbout reading pace	1.9M	Open →
Qwen3.6-27BQwen · 28B params	Q4_K_M	18.6 GB	5.45 GB	~14 tok/sAbout reading pace	2.5M	Open →
Qwen3.5-2BQwen · 2.3B params	Q4_K_M	2.32 GB	21.7 GB	~165 tok/s	4.9M	Open →
gemma-4-E4B-itGoogle · 8.0B params	Q4_K_M	5.86 GB	18.1 GB	~67 tok/s	4.4M	Open →
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA · 32B params	Q4_K_M	20.3 GB	3.71 GB	~10 tok/sAbout reading pace	530K	Open →
NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA · 4.0B params	Q4_K_M	3.46 GB	20.5 GB	~77 tok/s	3.4M	Open →
gemma-4-E2B-itGoogle · 5.1B params	Q4_K_M	4.02 GB	20.0 GB	~178 tok/s	3M	Open →
Qwen3.5-0.8BQwen · 873M params	Q4_K_M	1.46 GB	22.5 GB	~362 tok/s	2.6M	Open →
Qwen3-VL-8B-InstructQwen · 8.8B params	Q4_K_M	7.04 GB	17.0 GB	~35 tok/s	14.6M	Open →
Qwen3.5-27BQwen · 28B params	Q4_K_M	18.6 GB	5.45 GB	~14 tok/sAbout reading pace	1.9M	Open →
North-Micro-Vision-InstructCohereLabs · 2.5B params	Q4_K_M	2.99 GB	21.0 GB	~109 tok/s	180.7K	Open →
Qwen3.5-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	0.90 GB	~110 tok/s	1.6M	Open →
granite-4.2-8bIBM · 8.8B params	Q4_K_M	7.49 GB	16.5 GB	~32 tok/s	128.4K	Open →
GLM-4.7-Flashzai-org · 31B params	Q4_K_M	20.4 GB	3.59 GB	~10 tok/sAbout reading pace	1.8M	Open →
granite-4.1-3bIBM · 3.4B params	Q4_K_M	3.72 GB	20.3 GB	~74 tok/s	521.8K	Open →
LFM2.5-2.6BLiquidAI · 2.7B params	Q4_K_M	2.61 GB	21.4 GB	~118 tok/s	108.1K	Open →
Qwen3-VL-4B-InstructQwen · 4.4B params	Q4_K_M	4.72 GB	19.3 GB	~55 tok/s	3.5M	Open →
granite-4.1-30bIBM · 29B params	Q4_K_M	20.4 GB	3.56 GB	~10 tok/sAbout reading pace	301.5K	Open →
Qwen3-VL-2B-InstructQwen · 2.1B params	Q4_K_M	3.05 GB	21.0 GB	~100 tok/s	2.8M	Open →
granite-4.2-3bIBM · 3.7B params	Q4_K_M	3.72 GB	20.3 GB	~74 tok/s	50.6K	Open →
LFM2.5-230MLiquidAI · 230M params	Q4_K_M	1.05 GB	22.9 GB	~823 tok/s	88.6K	Open →
Qwen3-0.6BQwen · 752M params	Q4_K_M	2.22 GB	21.8 GB	~153 tok/s	29.7M	Open →
gpt-oss-20bOpenAI · 21B params	Q4_K_M	13.8 GB	10.2 GB	~82 tok/s	6.6M	Open →
granite-4.2-30bIBM · 29B params	Q4_K_M	20.7 GB	3.33 GB	~10 tok/sAbout reading pace	35.8K	Open →
granite-4.1-8bIBM · 8.8B params	Q4_K_M	7.49 GB	16.5 GB	~32 tok/s	179.3K	Open →
NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA · 32B params	Q4_K_M	20.3 GB	3.71 GB	~10 tok/sAbout reading pace	875.2K	Open →
LFM2.5-VL-3BLiquidAI · 3.1B params	Q4_K_M	2.85 GB	21.1 GB	~118 tok/s	26.7K	Open →
MiMo-V2.6-Distill-Qwen-9BNEWXiaomiMiMo · 9.4B params	Q4_K_M	6.90 GB	17.1 GB	~42 tok/s	14.6K	Open →
Qwen3-4B-Instruct-2507Qwen · 4.0B params	Q4_K_M	4.72 GB	19.3 GB	~55 tok/s	3.7M	Open →
Ling-3.0-tinyinclusionAI · 7.9B params	Q4_K_M	5.85 GB	18.1 GB	~238 tok/s	17.7K	Open →
Qwen3-8BQwen · 8.2B params	Q4_K_M	7.04 GB	17.0 GB	~35 tok/s	10.7M	Open →
Olmo-3-7B-Instructallenai · 7.3B params	Q4_K_M	7.96 GB	16.0 GB	~29 tok/sAbout reading pace	481.7K	Open →
Qwen3-4BQwen · 4.0B params	Q4_K_M	4.51 GB	19.5 GB	~55 tok/s	7.8M	Open →
LFM2.5-8B-A1BLiquidAI · 8.5B params	Q4_K_M	6.06 GB	17.9 GB	~184 tok/s	32.4K	Open →

Sized at the format most people actually download, not at FP16. The catalogue ordered by downloads →

Rent it instead

Rent an L4 by the hour

What the NVIDIA L4 costs on Vast.ai and RunPod right now, per machine, from one card to eight. Every price is read from the provider's own API.

Prices · 18:30 UTC, 22 Sept

1× L4 · 24 GB

$0.26per hour

Vast.ai · Marketplace · 98.9% reliableRunPod $0.49/h Secure CloudRent on Vast.ai

2× L4 · 48 GB

$0.65per hour

Vast.ai · Marketplace · 99.4% reliableRunPod $0.98/h Secure CloudRent on Vast.ai

4× L4 · 96 GB

$1.29per hour

Vast.ai · Marketplace · 99.4% reliableRent on Vast.ai

Buy one or rent one?

Your price, your hours, your electricity. Everything else is arithmetic.

Price you would pay ($)

Resale after 3 years ($)

Hours of use per day4 h

$ per kWh

Watts

Rent $/h

Watts start at the NVIDIA L4’s published board power (72 W), an upper bound: decoding rarely holds a card at its limit. Rent starts at the cheapest live price (Vast.ai). The electricity price is a placeholder — put yours in.

At 4 h a day, renting costs $379.60 a year.

That is $31.63 a month, and nothing when the machine is stopped.
Owning would cost $26.28 a year in electricity, on top of the price.
Every $1,000 of purchase price takes 2.8 years of this use to earn back.

Enter the price you would pay to see the exact break-even.

On-demand prices for the whole machine. Vast hosts below 98% measured reliability are left out; RunPod’s Community Cloud is vetted third-party hosts and its Secure Cloud is data-centre capacity. Every rentable card compared.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Under the hood

The specification behind every figure

What the manufacturer publishes for this device, and the pages it was read from.

Manufacturer specification

Memory scope	dedicated
Capacity	24 GB
Published options	24 GB
Bandwidth	300 GB/s
Memory type	Not published
Bus width	Not published
FP32 peak	Not published
Dense matrix peak	Not published without sparsity
TDP	72 W

Source ledger

NVIDIA L4 Tensor GPU ↗

Retrieved 2026-09-13

Caveats

The page's Tensor Core throughput figures are footnoted "Shown with sparsity" and are not used.
A passively cooled data-center card: the TDP assumes server airflow.
No dense matrix throughput is published for this device in a form this catalogue accepts, so no compute roof is priced and time to first token is withheld.

Published capacity is a hardware ceiling, not guaranteed free runtime memory. The calculator shows the runtime reserve separately rather than folding it into a single number.

Run the diagnostic on the NVIDIA L4 →