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

Microsoft · Downloadable

Mage-VL

VisionVideo understanding

Understands text and images and responds in text.

Source listed
Jul 25, 2026
Parameters
4.7B
4-bit weights only
≈ 2.4 GB
License
apache-2.0
Compare this model →

Baidu · Downloadable

Qianfan-OCR

VisionDocument reading

Understands text and images and responds in text.

Source listed
Mar 18, 2026
Parameters
4.7B
4-bit weights only
≈ 2.4 GB
License
apache-2.0
Compare this model →

BAAI · Downloadable

AREX-Turbo

Vision

Understands text and images and responds in text.

Source listed
Jul 23, 2026
Parameters
4.5B
4-bit weights only
≈ 2.3 GB
License
apache-2.0
Compare this model →

Tencent · Downloadable

POINTS-Reader

Vision

Understands text and images and responds in text.

Source listed
Aug 15, 2025
Parameters
4.1B
4-bit weights only
≈ 2.1 GB
License
apache-2.0
Compare this model →

OpenBMB · Downloadable

MiniCPM-V-4

VisionDocument readingEmbeddings

Understands text and images and responds in text.

Source listed
Jul 12, 2025
Parameters
4.1B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

Tencent · Downloadable

TCAndon-Router

Text generation

Generates text for conversations and writing.

Source listed
Jan 7, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

BAAI · Downloadable

Bunny-v1_0-4B

Text generation

Generates text for conversations and writing.

Source listed
Apr 30, 2024
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

BAAI · Downloadable

Bunny-v1_1-4B

Text generation

Generates text for conversations and writing.

Source listed
May 8, 2024
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

inclusionAI · Downloadable

DR-Venus-4B-RL

Text generation

Generates text, according to its published model architecture (Qwen3ForCausalLM). No task is listed on Hugging Face, so check the model card for intended use.

Source listed
Apr 21, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
Not disclosed
Compare this model →

inclusionAI · Downloadable

DR-Venus-4B-SFT

Text generation

Generates text, according to its published model architecture (Qwen3ForCausalLM). No task is listed on Hugging Face, so check the model card for intended use.

Source listed
Apr 21, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
Not disclosed
Compare this model →

inclusionAI · Downloadable

VISTA-4B

VisionReinforcement learning

Understands text and images and responds in text.

Source listed
Jun 12, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
Compare this model →

inclusionAI · Downloadable

ZwZ-4B

Vision

Understands text and images and responds in text.

Source listed
Feb 12, 2026
Parameters
4B
4-bit weights only
≈ 2 GB
License
apache-2.0
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