Compatibility directory

GPU × local AI model compatibility

Explore dedicated-VRAM fit between selected NVIDIA RTX GPUs and local AI models. Each compatibility page uses the same deterministic memory-planning engine as AIPCFit instead of separate hand-written thresholds.

44

Compatibility checks

3

Representative GPUs

1

Shared fit engine

GPU LLM compatibility

Check GPU and LLM Compatibility

Choose a GPU and compare which local LLMs fit its memory constraints and compatibility assumptions. Each listing stays centered on a specific GPU and model pairing using AIPCFit's deterministic memory-planning data.

GPU/model pairs

Browse Compatibility by GPU

12GB VRAM · Ampere

RTX 3060 12GB

13 model-level compatibility checks currently meet AIPCFit's publication threshold for this GPU.

Open GPU overview →

Meta

Llama 3.2 3B

good fit
2 recorded variantsView compatibility →

Meta

Llama 3.1 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 4B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 14B

fits
2 recorded variantsView compatibility →

Google

Gemma 3 4B

good fit
2 recorded variantsView compatibility →

Google

Gemma 3 12B

good fit
2 recorded variantsView compatibility →

Mistral AI

Mistral 7B

good fit
2 recorded variantsView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 7B

good fit
2 recorded variantsView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 14B

good fit
1 recorded variantView compatibility →

Microsoft

Phi-3.5 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4

good fit
1 recorded variantView compatibility →

16GB VRAM · Blackwell

RTX 5060 Ti 16GB

13 model-level compatibility checks currently meet AIPCFit's publication threshold for this GPU.

Open GPU overview →

Meta

Llama 3.2 3B

good fit
2 recorded variantsView compatibility →

Meta

Llama 3.1 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 4B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 14B

good fit
2 recorded variantsView compatibility →

Google

Gemma 3 4B

good fit
2 recorded variantsView compatibility →

Google

Gemma 3 12B

good fit
2 recorded variantsView compatibility →

Mistral AI

Mistral 7B

good fit
2 recorded variantsView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 7B

good fit
2 recorded variantsView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 14B

good fit
1 recorded variantView compatibility →

Microsoft

Phi-3.5 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4

good fit
1 recorded variantView compatibility →

24GB VRAM · Ada Lovelace

RTX 4090 24GB

18 model-level compatibility checks currently meet AIPCFit's publication threshold for this GPU.

Open GPU overview →

Meta

Llama 3.2 3B

good fit
2 recorded variantsView compatibility →

Meta

Llama 3.1 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 4B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 8B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 14B

good fit
2 recorded variantsView compatibility →

Alibaba Cloud

Qwen3 30B-A3B

fits
1 recorded variantView compatibility →

Alibaba Cloud

Qwen3.8 27B

fits
1 recorded variantView compatibility →

Google

Gemma 3 4B

good fit
2 recorded variantsView compatibility →

Google

Gemma 3 12B

good fit
2 recorded variantsView compatibility →

Google

Gemma 3 27B

good fit
2 recorded variantsView compatibility →

Mistral AI

Mistral 7B

good fit
2 recorded variantsView compatibility →

Mistral AI

Mistral Small 3.2 24B

good fit
1 recorded variantView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 7B

good fit
2 recorded variantsView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 14B

good fit
1 recorded variantView compatibility →

DeepSeek

DeepSeek-R1 Distill Qwen 32B

fits
1 recorded variantView compatibility →

Microsoft

Phi-3.5 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4 Mini

good fit
1 recorded variantView compatibility →

Microsoft

Phi-4

good fit
1 recorded variantView compatibility →

Model directory

Browse local AI models

Compare model families, sizes, quantizations, and memory-planning data.

GPU directory

Browse NVIDIA GPUs

Explore GPU overviews, VRAM tiers, and runtime-specific guidance.

Hardware fit

What LLM can I run?

Start from your hardware instead of a specific GPU-model pair.

What these checks mean

Compatibility here is a memory-planning signal, not a benchmark or speed guarantee. Context length, runtime overhead, system RAM, CPU offload, drivers, and software configuration can change real-world behavior.

Read the methodology →