Skip to content
24GB VRAM

RTX 5090 Laptop GPU — what LLMs can it run?

119 of 151 indexed models fit comfortably in 24GB at 4K context, each at the highest-quality quant that still leaves headroom.

Context length
Concurrency (batch)

The short answer for RTX 5090 Laptop GPU

24 GB, GDDR7, bandwidth not listed. 119 of 151 models in this index fit comfortably at 4K context; 124 load at all.

Hardware specifications source
Biggest that fits
Kolibri-1 78B-A3.5B — 20.9 GB at IQ1_M — 3.1 GB spare at 4K, so longer context comes out of a thin margin.Download: bartowski/Kolibri-1-GGUF · Kolibri-1-IQ1_M.gguf · 19.77 GB · listing read 2026-11-12
Room to grow
Ternary Bonsai 2 27B — 8.2 GB at Q2_0 — under 60% of the card, which leaves 15.8 GB for a long context window or a second process.Download: prism-ml/Ternary-Bonsai-2-27B-gguf · Ternary-Bonsai-2-27B-Q2_0.gguf · 7.63 GB · + vision projector 0.63 GB, separate · listing read 2026-11-19
The wall you will hit
Laptop cooling and power limits affect performance even between laptops with the same GPU. Use the listed VRAM for fit, not desktop benchmark results for speed.
Best local LLM for 24GB

50B+ · 2

RTX 5090 Laptop GPU — what LLMs can it run? — 50B+
ModelQuantEst. VRAMHeadroom
Kolibri-1 78B-A3.5B78B-A3.5BIQ1_M20.9 GB+3.1 GB
Ripple-65B-Preview65BIQ1_M19.98 GB+4 GB

20–50B · 39

RTX 5090 Laptop GPU — what LLMs can it run? — 20–50B
ModelQuantEst. VRAMHeadroom
Recursive Flint 40B-A2.5B40B-A2.5BQ3_K_M20.6 GB+3.4 GB
Seed-OSS 36B Instruct36BQ3_K_M19.52 GB+4.5 GB
Apex 35B V1.035B-A3BQ3_K_M18.71 GB+5.3 GB
Command R 35B35BGPTQ INT418.97 GB+5 GB
MiniCPM-V 4.7 35B-A3B35B-A3BQ4_K_S20.88 GB+3.1 GB
Clover-35B-A3B-Preview35B-A3BQ4_020.91 GB+3.1 GB
Yi 1.5 34B Chat34BAWQ INT419 GB+5 GB
Nanbeige4.3 33B33BQ3_K_M17.55 GB+6.5 GB
Qwen3 32B Instruct32BAWQ INT418.22 GB+5.8 GB
Qwen2.5 32B Instruct32BEXL2 3.5bpw15.96 GB+8 GB
Qwen2.5-Coder 32B Instruct32BAWQ INT418.08 GB+5.9 GB
DeepSeek-R1-Distill-Qwen-32B32BEXL2 3.5bpw15.96 GB+8 GB
Nemotron 3.5 Lightning 30B-A3B30B-A3BQ4_019.79 GB+4.2 GB
Xing4.0 29B-A4B29B-A4BIQ4_XS19 GB+5 GB
Gemma 4 31B IT31BQ4_020.08 GB+3.9 GB
Qwen3-Coder 30B-A3B Instruct30B-A3BQ4_K_M19.81 GB+4.2 GB
Qwen3-VL 30B-A3B Instruct30B-A3BQ4_K_M19.89 GB+4.1 GB
Qwen3 30B-A3B Instruct30B-A3BQ4_K_M19.73 GB+4.3 GB
North Mini Code 1.0 30B-A3B30B-A3BQ4_K_M20.01 GB+4 GB
Hy-Memory-30B-A3B30B-A3BQ4_K_M19.46 GB+4.5 GB
GLM-4.7-Flash30B-A3BQ4_K_M19.51 GB+4.5 GB
Granite 4.2 30B30BQ4_K_M19.61 GB+4.4 GB
Muse Glimmer 30B30BQ4_K_M17.66 GB+6.3 GB
Qwen3.8 27B27BQ4_K_M18.49 GB+5.5 GB
Arete 27B27BQ4_K_M18.25 GB+5.8 GB
Qwen3.6 27B27BQ4_K_M19.05 GB+4.9 GB
DaisyRose 27B27BQ4_K_M18.38 GB+5.6 GB
Gemma 2 27B Instruct27BQ4_K_M18.81 GB+5.2 GB
Gemma 3 27B IT27BQ4_K_M18.3 GB+5.7 GB
Ternary Bonsai 2 27B27BQ2_08.24 GB+15.8 GB
Clairic 27B27BQ5_K_M20.51 GB+3.5 GB
Gemma 4 26B-A4B IT26B-A4BQ4_K_M18.22 GB+5.8 GB
Mistral Small 24B Instruct24BEXL2 4.65bpw15.26 GB+8.7 GB
LFM2.5 24B-A2B24B-A2BQ4_K_M15.27 GB+8.7 GB
Devstral Small 1.1 24B24BQ6_K20.89 GB+3.1 GB
Magistral Small 1.2 24B24BQ6_K20.89 GB+3.1 GB
Codestral 22B22BQ4_K_M15.03 GB+9 GB
ERNIE 4.5 21B-A3B21B-A3BQ4_K_M14.18 GB+9.8 GB
GPT-OSS 20B21B MoEMXFP412.76 GB+11.2 GB

10–20B · 19

RTX 5090 Laptop GPU — what LLMs can it run? — 10–20B
ModelQuantEst. VRAMHeadroom
InternLM2 20B Chat20BQ5_K_M15.59 GB+8.4 GB
DeepSeek-Coder-V2-Lite Instruct16B-A2.4BQ8_017.56 GB+6.4 GB
DeepSeek-V2-Lite Chat16B-A2.4BQ4_K_M10.08 GB+13.9 GB
StarCoder2 15B15BQ4_K_M9.85 GB+14.2 GB
Qwen3 14B Instruct14BQ5_K_M11.65 GB+12.4 GB
Qwen2.5 14B Instruct14BQ5_K_M11.73 GB+12.3 GB
DeepSeek-R1-Distill-Qwen-14B14BEXL2 4.65bpw9.75 GB+14.3 GB
Phi-4 14B14BQ5_K_M11.77 GB+12.2 GB
Phi-3 Medium 14B Instruct14BQ6_K12.86 GB+11.1 GB
Ministral 3 14B Instruct14BQ8_015.7 GB+8.3 GB
Mistral Nemo 12B Instruct12BQ6_K11.14 GB+12.9 GB
Gemma 3 12B IT12BQ5_K_M9.84 GB+14.2 GB
Mellum2.1 12B-A2.5B Thinking12B-A2.5BQ8_013.6 GB+10.4 GB
Stable LM 2 12B Chat12BQ4_K_M8.52 GB+15.5 GB
Gemma 4 12B IT12BQ8_013.82 GB+10.2 GB
Llama 3.2 11B Vision Instruct11BQ8_012.9 GB+11.1 GB
Solar 10.7B Instruct11BQ4_K_M7.6 GB+16.4 GB
Falcon 3 10B Instruct10BQ4_K_M7.28 GB+16.7 GB
Metis-1.6 Logos10B MoEQ8_011.67 GB+12.3 GB

4–10B · 37

RTX 5090 Laptop GPU — what LLMs can it run? — 4–10B
ModelQuantEst. VRAMHeadroom
Gemma 2 9B Instruct9BQ8_011.7 GB+12.3 GB
Qwen3.5 9B9BQ8_010.39 GB+13.6 GB
K2 Horizon 7B7BQ8_010.62 GB+13.4 GB
GLM-4-9B-Chat9BQ8_010.16 GB+13.8 GB
MiMo-V2.6-Distill-Qwen-9B9BQ8_010.11 GB+13.9 GB
Clairic Flash 9B9BQ8_010.1 GB+13.9 GB
Qwen3-VL 8B Instruct8BQ8_010.39 GB+13.6 GB
TwIL 8B balanced8.8BQ4_K_M6.28 GB+17.7 GB
Granite 4.2 8B8BQ8_010.46 GB+13.5 GB
Ministral 3 8B Instruct8BQ8_010.02 GB+14 GB
LFM2.5 8B-A1B8B-A1.5BQ8_09.46 GB+14.5 GB
Qwen2-VL 7B Instruct7BQ4_K_M5.49 GB+18.5 GB
Granite 3.1 8B Instruct8BQ4_K_M5.88 GB+18.1 GB
Shot Reader 8B8BQ8_09.72 GB+14.3 GB
Qwen3 8B Instruct8BQ6_K7.64 GB+16.4 GB
Llama 3.1 8B Instruct8BQ8_09.47 GB+14.5 GB
Nous Hermes 3 Llama 3.1 8B8BEXL2 4.65bpw5.43 GB+18.6 GB
Aya 23 8B8BQ4_K_M5.64 GB+18.4 GB
OpenChat 3.6 8B8BEXL2 4.65bpw5.43 GB+18.6 GB
DeepSeek-R1-Distill-Llama-8B8BQ5_K_M6.51 GB+17.5 GB
InternLM2 7B Chat7BQ4_K_M5.45 GB+18.6 GB
Qwen2.5 7B Instruct7BQ6_K6.77 GB+17.2 GB
Qwen2.5-Coder 7B Instruct7BEXL2 4.65bpw4.87 GB+19.1 GB
DeepSeek-R1-Distill-Qwen-7B7BEXL2 4.65bpw4.87 GB+19.1 GB
Gemma 4 E4B ITE4BQ8_08.68 GB+15.3 GB
Nemotron 3.5 Bolt E4BE4BQ8_08.5 GB+15.5 GB
OLMo 2 7B Instruct7BQ8_010.3 GB+13.7 GB
OLMo 3 7B Instruct7BQ8_010.3 GB+13.7 GB
WizardLM-2 7B7BQ4_K_M5.14 GB+18.9 GB
Mistral 7B Instruct v0.37BQ6_K6.75 GB+17.3 GB
Zephyr 7B Beta7BQ6_K6.75 GB+17.3 GB
K2 Horizon 3.7B3.7BQ8_06.24 GB+17.8 GB
Gemma 4 E2B ITE2BQ8_05.2 GB+18.8 GB
Qwen3.5 4B4BQ8_04.97 GB+19 GB
Gemma 3 4B IT4BQ8_05.05 GB+19 GB
Spark-X2.6 4B4BQ8_04.84 GB+19.2 GB
Qwen3 4B Instruct4BQ6_K4.05 GB+20 GB

<4B · 22

RTX 5090 Laptop GPU — what LLMs can it run? — <4B
ModelQuantEst. VRAMHeadroom
Phi-4 Mini Instruct3.8BQ8_04.81 GB+19.2 GB
Phi-3.5 Mini Instruct3.8BQ8_05.89 GB+18.1 GB
Granite 4.2 3B3BQ8_04.41 GB+19.6 GB
Ministral 3 3B Instruct3BQ8_04.26 GB+19.7 GB
Llama 3.2 3B Instruct3BQ8_04.05 GB+20 GB
Qwen2.5 3B Instruct3BQ8_03.59 GB+20.4 GB
SmolLM3 3B3BQ8_03.73 GB+20.3 GB
Metis-1.6 Praxis3B MoEQ8_03.4 GB+20.6 GB
Gemma 2 2B Instruct2BQ8_03.34 GB+20.7 GB
Qwen3.5 2B2BQ8_02.23 GB+21.8 GB
LFM3 1.7B1.7BQ8_01.99 GB+22 GB
LFM3-VL 2B2BQ8_01.99 GB+22 GB
Qwen3 1.7B Instruct1.7BQ8_02.39 GB+21.6 GB
Qwen2.5 1.5B Instruct1.5BQ8_01.83 GB+22.2 GB
Llama 3.2 1B Instruct1BQ8_01.51 GB+22.5 GB
LFM2.5 1.2B Instruct1.2BQ8_01.35 GB+22.7 GB
K2 Horizon 0.9B0.9BQ8_01.44 GB+22.6 GB
Qwen3.5 0.8B0.8BQ8_00.94 GB+23.1 GB
BotPsy 0.8B0.75BQ8_00.9 GB+23.1 GB
Qwen2.5 0.5B Instruct0.5BQ8_00.6 GB+23.4 GB
MNN-Pocket 0.3B Instruct0.36BQ8_00.46 GB+23.5 GB
MNN-Pocket 0.3B QAT0.30BQ4_10.26 GB+23.7 GB

How this list is built

Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so each estimate leaves headroom. Raise the context length and the list shortens; the calculator lets you check any combination directly.

A further 5 models load but with no headroom to spare (up to 105% of VRAM) — the Quant Hub’s GPU chips count those too, which is why its number is higher.

Measured on this card

No inference runs were measured on this laptop GPU. Fit figures are estimates, not measured results.

Cards with the same budget

What fits is decided by memory, so every 24GB card of this type returns the same list. These pages are not different answers — they differ in throughput, which this index does not measure per card.

Setup guides for this card

Step-by-step guides written for the RTX 5090 Laptop GPU's runtime or memory size, with the commands and the checks that prove the model is running on the card.

Stepping up

A RTX 5090 (32GB) fits 6 more of the indexed models than this card. RTX 5090 →

Common questions

What is the best local LLM for an RTX 5090 Laptop GPU?

For everyday use, Ternary Bonsai 2 27B at Q2_0 — about 8.2 GB of the card's 24 GB at 4K context, leaving 15.8 GB for a longer window. If you want the largest thing that will load, that is Kolibri-1 78B-A3.5B at IQ1_M (20.9 GB, 3.1 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.

Can an RTX 5090 Laptop GPU run Llama 3.2 90B Vision Instruct?

No. Its smallest build here, Q3_K_M, needs about 46.2 GB and this card has 24.0 GB usable — short by 22.2 GB before any context beyond 4K. The largest model this card does clear is Kolibri-1 78B-A3.5B.

RTX 5090 Laptop GPU or RTX 4090 for local LLMs?

They hold the same models: 24 GB against 24 GB fits 119 and 119 of 151 respectively at 4K. Bandwidth is not recorded for both cards here, so this page will not rank them on speed.

119 of 151 indexed models fit comfortably in 24GB at 4K context, each at the highest-quality quant that still leaves headroom.

This page's figures change when the model or the runtime does.Last updated 2026-11-22 RSS → /feed.xml