YOUR MEMORY BUDGET

Local AI models for 12GB VRAM.

Explore models by theoretical weight storage. This shortlist covers available downloadable language models with a known or name-derived parameter count.

10 GBleft for weights at 4-bit
12 GB budget − 2 GB reserve

The reserve is your assumption, not a measured overhead allowance. The default 2 GB may be too little for your runtime or context length. Results are not confirmed fits and have no verified speed rating. Weight estimates use decimal GB and do not include quantization overhead. For timed results on one 12GB machine, see measured speeds.

978 models within the weight budget

Largest parameter count first · not a quality ranking

Qwen · Alibaba · Downloadable

Qwen3-4B

Text generation

Generates text for conversations and writing.

Source listed
Apr 27, 2025
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Qwen · Alibaba · Downloadable

Qwen3-4B-SafeRL

Text generation

Generates text for conversations and writing.

Source listed
Sep 30, 2025
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Qwen · Alibaba · Downloadable

Qwen3-4B-Thinking-2507

Text generationReasoning

Generates text for conversations and writing. Supports a reasoning mode.

Source listed
Aug 5, 2025
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Qwen · Alibaba · Downloadable

Qwen3-VL-4B-Thinking

VisionReasoning

Understands text and images and responds in text. Supports a reasoning mode.

Source listed
Oct 11, 2025
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Qwen · Alibaba · Downloadable

Qwen3.5-4B

Vision

Understands text and images and responds in text.

Source listed
Feb 27, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Qwen · Alibaba · Downloadable

Qwen3.5-4B-Base

Vision

Understands text and images and responds in text.

Source listed
Feb 27, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Tencent · Downloadable

WeVisOne-VL-4B

VisionDocument readingObject detection

Understands text and images and responds in text.

Source listed
Oct 22, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Microsoft · Downloadable

Phi-4-mini-reasoning

Text generationReasoning

Generates text for conversations and writing. Supports a reasoning mode.

Source listed
Apr 29, 2025
Parameters
3.8B
4-bit weights only
≈ 1.9 GB
License
mit
Compare this model →

Microsoft · Downloadable

MediPhi

Text generation

Generates text for conversations and writing.

Source listed
Feb 3, 2025
Parameters
3.8B
4-bit weights only
≈ 1.9 GB
License
mit
Compare this model →

Microsoft · Downloadable

MediPhi-MedCode

Text generationCoding

Designed for coding and software tasks. Generates text for conversations and writing.

Source listed
May 29, 2025
Parameters
3.8B
4-bit weights only
≈ 1.9 GB
License
mit
Compare this model →

What changes with a 12GB budget?

With 2 GB set aside, 10 GB remains for weights. At 4-bit that allows up to 20 billion parameters in the theoretical calculation. An 8-billion-parameter model alone needs approximately 4 GB at this precision, before its conversation cache and runtime.

Why might a listed model still fail to load?

Its actual weight format may need more space, the runtime may not support its architecture, or the conversation cache may exceed your reserve. Vision models can also need space for image processing. Check the original model card and measure your exact setup before relying on it.

Read the memory guide and technical sources · Compare two models