What AI models can you run with 12GB of VRAM?

With 12GB of VRAM, 170 of 225 catalog builds have good/excellent reference fits at 4K context. 32 more run by offloading part of the model to system RAM, which works but is slower.

170

run well

with trade-offs

225

builds evaluated

Check my actual PC

Compare a larger model with a longer context

A 12 GB budget is shared by cards such as the RTX 3060 and RTX 4070. After finding a build that fits, compare spending the remaining memory on context versus a higher-bit quantization of the same model.

Quantization compresses weights; it does not remove context-cache or runtime costs. Compare the same model at Q4 and Q8 in the calculator, then validate its output on your task. Training and fine-tuning need a different memory budget.

Change system RAM and context Compare exact builds 8 GB vs 12 GB vs 16 GB

Four builds to compare by task

Editorial examples chosen for their documented tasks, all Q4_K_M at 4,096 tokens. Check the verdict before downloading. These are estimates, not benchmark rankings.

Short conversations and summaries

llama3.2:3b

A small instruction-tuned model for trying local chat with more memory headroom. Review summaries against the source text.

Task and build source · Catalog checked 2026-10-13

Excellent fit

Est. memory 2.9 GB · high confidence

Model fully fits in usable GPU VRAM.

ollama run llama3.2:3b

General instruction following

llama3.1:8b

An 8B instruction-tuned alternative to compare on your own prompts. It needs more memory than the 3B example and can still give incorrect answers.

Task and build source · Catalog checked 2026-10-13

Excellent fit

Est. memory 5.8 GB · high confidence

Model fully fits in usable GPU VRAM.

ollama run llama3.1:8b

Code explanation and editing

qwen2.5-coder:7b

A code-focused instruct build. Review generated changes and run your tests; a memory fit is not evidence that its code is correct.

Task and build source · Catalog checked 2026-10-13

Excellent fit

Est. memory 5.3 GB · high confidence

Model fully fits in usable GPU VRAM.

ollama run qwen2.5-coder:7b

A larger general-purpose comparison

qwen2.5:14b

A 14B instruct build for comparing answers when memory allows. Larger size does not establish better results for every task.

Task and build source · Catalog checked 2026-10-13

Good fit

Est. memory 10.2 GB · high confidence

Model fully fits in usable GPU VRAM.

ollama run qwen2.5:14b

After opening Ollama's prompt, enter /set parameter num_ctx 4096 to match this calculation, within the model's supported maximum. Install a current Ollama release that supports the build. Setup and troubleshooting.

Models that fit — largest memory first — for 12GB of VRAM

Heaviest first — each of these has a good/excellent reference fit with 12GB at 4K context, counting weights, KV cache, compute buffers and runtime overhead.

Models that fit in 12GB of VRAM
Model	Build	Memory needed	Verdict
everythinglm	everythinglm:13b-16k-q4_K_MQ4_K_M	~11.4 GB	Good fit
nexusraven	nexusraven:13b-q4_K_MQ4_K_M	~11.4 GB	Good fit
open-orca-platypus2	open-orca-platypus2:13b-q4_K_MQ4_K_M	~11.4 GB	Good fit
codeup	codeup:13b-llama2-chat-q4_K_MQ4_K_M	~11.4 GB	Good fit
wizard-vicuna	wizard-vicuna:13b-q4_K_MQ4_K_MMemory scenario exceeds documented maximum of 2,048 tokens.	~11.4 GB	Good fit
wizardlm-uncensored	wizardlm-uncensored:13b-llama2-q4_K_MQ4_K_M	~11.4 GB	Good fit
clef-flash	clef-flash:9bQ8_0	~11.0 GB	Good fit
phi4	phi4:14bQ4_K_M	~10.3 GB	Good fit
llama3.2-vision	llama3.2-vision:11bQ4_K_M	~8.9 GB	Good fit
mistral-nemo	mistral-nemo:12b-instruct-2407-q4_K_MQ4_K_M	~8.5 GB	Excellent fit
falcon2	falcon2:11b-q4_K_MQ4_K_MMemory scenario exceeds documented maximum of 2,048 tokens.	~8.2 GB	Excellent fit
llama-pro	llama-pro:8b-text-q4_K_MQ4_K_M	~8.0 GB	Excellent fit
nous-hermes2	nous-hermes2:10.7b-solar-q4_K_MQ4_K_M	~7.6 GB	Excellent fit
solar	solar:10.7b-text-v1-q4_K_MQ4_K_M	~7.6 GB	Excellent fit
deepseek-ocr	deepseek-ocr:3bF16	~7.4 GB	Excellent fit
olmo2	olmo2:7bQ4_K_M	~6.9 GB	Excellent fit
olmo-3	olmo-3:7b-instruct-q4_K_MQ4_K_M	~6.9 GB	Excellent fit
codegeex4	codegeex4:9b-all-q4_K_MQ4_K_M	~6.8 GB	Excellent fit
glm4	glm4:9b-text-q4_K_MQ4_K_M	~6.8 GB	Excellent fit
nimble-v3	nimble-v3:9b-q4_K_MQ4_K_M	~6.6 GB	Excellent fit
ornith-1.5	ornith-1.5:9bQ4_K_M	~6.6 GB	Excellent fit
nimble	nimble:9b-q4_K_MQ4_K_M	~6.6 GB	Excellent fit
ornith	ornith:9bQ4_K_M	~6.6 GB	Excellent fit
nemotron-3-nano	nemotron-3-nano:4bQ4_K_M	~6.5 GB	Excellent fit
deepseek-llm	deepseek-llm:7b-base-q4_K_MQ4_K_M	~6.5 GB	Excellent fit

Graphics cards with 12GB of VRAM

GeForce RTX 5070 GeForce RTX 4070 Ti GeForce RTX 4070 GeForce RTX 3060 12GB Radeon RX 6700 XT

How this was calculated

Every model is evaluated at 4K context against 12GB of VRAM, minus a reserve for the display and operating system, and assuming 32GB of system RAM with an illustrative 8-core/16-thread CPU and 500GB free disk. Available RAM and GPU budget/free memory are unknown. A working Vulkan backend on Windows is assumed; this is a capacity scenario, not a claim that all cards with this much memory have the same support. Your own machine may differ, which is what the PC scan measures directly.

12GB VRAM and local AI — frequently asked questions

Is 12GB of VRAM enough to run AI models locally?: Yes — 170 of the 225 builds evaluated have good/excellent reference fits with 12GB at 4K context, and 32 more run with part of the model offloaded to system RAM. The heaviest that fits is everythinglm:13b-16k-q4_K_M, needing about 11.4 GB.
How should I choose a model for 12GB VRAM?: Choose for your task, then check memory. everythinglm (everythinglm:13b-16k-q4_K_M) is the heaviest catalog default that fits this reference budget; its size does not establish answer quality. Smaller models leave headroom for a longer context window, which often matters more than raw parameter count.
Which graphics cards have 12GB of VRAM?: Cards with 12GB include GeForce RTX 5070, GeForce RTX 4070 Ti, GeForce RTX 4070, GeForce RTX 3060 12GB, Radeon RX 6700 XT.

Other VRAM amounts

4GB 6GB 8GB 10GB 16GB 24GB 32GB