12GB GPUs for local AI
The sweet spot for 8B-14B models with room for longer context. 12B models like Gemma 4 12B fit fully on the GPU.
Models that run well on this tier
gemma4:12b12B · Q4_K_M · ~9.5 GB
Newest Gemma at 12B; thinking + tools + vision.
qwen3.5:9b9B · Q4_K_M · ~8.1 GB
Excellent 8-12GB all-rounder.
ministral-3:14b14B · Q4_K_M · ~10.6 GB
12-16GB Mistral with long context.
qwen3:14b14B · Q4_K_M · ~10.8 GB
Good 12-16GB reasoning model.
qwen2.5-coder:14b14B · Q4_K_M · ~10.5 GB
12-16GB coding model.
phi414B · Q4_K_M · ~10.6 GB
Maths/logic strength; short 16K context.
Check your exact setup in the GPU picker → · Read the full guide →
Hardware picks
NVIDIA GeForce RTX 3060 (12GB)
12GB GDDR6 · bandwidth 360 GB/s
The classic budget local-AI card, cheap on the used market. Slower than newer cards but 12GB is what matters.
Check priceNVIDIA GeForce RTX 5070 (12GB)
12GB GDDR7 · bandwidth 672 GB/s
Fast memory gives roughly double the 3060's generation speed on the same models. Great gaming card too.
Check priceIntel Arc B580 (12GB)
12GB GDDR6 · bandwidth 456 GB/s
Cheapest new 12GB card. Runs via Vulkan in Ollama; check driver support for your OS first.
Check priceWe don't list prices because they change daily. Bandwidth figures are manufacturer specs. Disclosure: "Check price" links may be affiliate links (rel="sponsored"); we may earn a commission at no cost to you.
Get the Local AI weekly
New models worth running, what fits which GPU, and setup fixes. One short email a week, free. Unsubscribe any time.
Local-LLM App Starter
A tested starter for building your own app on Ollama: chat, RAG and a GPU-fit checker. Skip the setup hours.
kits.lilaforge.com → VibeEngineBest local LLM tools, ranked
Ollama, LM Studio, Open WebUI and llama.cpp scored from public evidence, with top picks.
vibeengine.lilaforge.com →