What can you run on a 24GB graphics card?

With 24GB VRAM you can run 20 of the 51 open-source models we track — up to about 35B parameters at Q4 quantization. The most popular that fits is Qwen3.8-27B, needing 19 GB and generating around ~40 tok/s on typical hardware of this size.

Typical machines: RTX 4090 · RTX 3090 · RX 7900 XTX. After the operating system takes its share, about 24 GB is free for a model.

Having the memory is not the same as having the speed

24GB VRAM decides whether a model loads. How fast it answers is decided by memory bandwidth — how quickly your machine can read the model, which it must do once for every word it writes. The table quotes a RTX 4090; pick yours below to see the difference. Why bandwidth and not the processor →

936 GB/sGeForce RTX 3090 (24GB)· NVIDIA published specification: 936.2 GB/s GDDR6X, 384-bit
960 GB/sRadeon RX 7900 XTX (24GB)· AMD published specification: 960 GB/s GDDR6, 384-bit
1008 GB/sGeForce RTX 4090 (24GB)· NVIDIA published specification: 1,008 GB/s GDDR6X, 384-bit

Models that run on 24GB VRAM in 2026

ModelSizeMemoryEst. speedDownloadsFeels like
LFM2.5-230MLiquid Ai230M808 MBRuns well~4271 tok/s58,011Fast
MiniCPM5-1BOpenBMB1.1B1.3 GBRuns well~1004 tok/s667,154Fast
MiniCPM5-2BOpenBMB2.5B2.3 GBRuns well~420 tok/s206,774Fast
LFM2.5-2.6BLiquid Ai2.7B2.5 GBRuns well~391 tok/s98,713Fast
LFM2.5-VL-3BLiquid Ai3.1B2.5 GBRuns well~391 tok/s31,296Fast
North-Micro-Vision-InstructCohere2.5B2.5 GBRuns well~437 tok/s120,354Fast
granite-4.2-3bIbm3.7B3.1 GBRuns well~292 tok/s32,017Fast
Mage-VLMicrosoft4.7B3.9 GBRuns well~241 tok/s29,690Fast
LFM2.5-8B-A1BLiquid Ai · MoE8.5B5.9 GBRuns well~1076 tok/s1B active80,632Fast
Ling-3.0-tinyInclusionai · MoE7.9B6.2 GBRuns well~679 tok/s1.6B active25,727Fast
granite-4.2-8bIbm8.8B6.6 GBRuns well~123 tok/s52,946Fast
Ornith-1.5-9BDeepreinforce9.7B6.9 GBRuns well~113 tok/s453,339Fast
gemma-4-12B-itGoogle DeepMind12B9.6 GBRuns well~91 tok/s2,958,440Fast
Qwen3.8-27BAlibaba / Qwen28B19 GBRuns well~40 tok/s7,703,400Fast
diffusiongemma-26B-A4B-itGoogle DeepMind · MoE26B19 GBRuns well~252 tok/s4B active708,393Fast
Muse-Glimmer-30BMeta AI30B20 GBRuns well~36 tok/s441,881Fast
granite-4.2-30bIbm29B21 GBTight~37 tok/s17,894Fast
North-Mini-Code-1.0Cohere · MoE30B22 GBTight~171 tok/s6.1B active11,742Fast
Laguna-XS-2.1Poolside · MoE33B23 GBTight~162 tok/s6.7B active32,445Fast
Nex-N2.5-miniNex Agi · MoE35B24 GBTight~155 tok/s7B active4,543Fast

Sorted by memory, smallest first — click any column to change it.

Upgrade

Including NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 (18B) and Ornith-1.5-35B-A3B (36B) — around ~275 tok/s on a typical machine that size.

See everything 32GB VRAM runs →

All figures at Q4_K_M quantization, 4K context. “Memory” includes the model, its context cache and runtime overhead. Speed is calculated, not benchmarked: 1008 GB/s × 0.65 ÷ bytes read per token, assuming a RTX 4090. Scale it by your own machine's bandwidth from the list above. Mixture-of-experts models read only their active parameters per token, which is why some large ones outrun smaller dense models.

What you'd need more memory for

The next models up, and what they ask for.

Other hardware

Verified against HuggingFace on 2026-09-15New to this? Start here →Runs on a laptop and up · last 120 days only