Hardware Lab
What AI can you run with 8 GB of VRAM?
A graphics card with 8 GB of memory runs 52 of the 59 tools in the Free Market that work on your own computer. 7 need a bigger card: vLLM, AI Toolkit, LTX-Video, Kohya's GUI, Wan 2.1, HunyuanVideo, and Mochi 1.
For language models, 8 GB holds a model of up to about 9 billion parameters at the common Q4 quality level, entirely on the card, which is the fast way to run it. Bigger models still run if you have enough main memory, but several times slower.
Check a specific model size Check my whole computer
Tools that need a graphics card and fit in 8 GB
Build image and video generation workflows by connecting blocks, with full control over every step.
- Some setup
- Windows · Mac · Linux
- Code
- Open source, GPL-3.0
- Models
- Open weights, your choice
- Runs
- On your computer
- Real cost
- Free
Fine-tune language models on your own data, faster and with much less memory.
- For tinkerers
- Windows · Linux
- Code
- Open source, Apache-2.0
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
New
Runs very large language models on an ordinary gaming PC by spreading them across the graphics card, memory and SSD.
- Some setup
- Windows · Mac · Linux
- Code
- Open source, MIT
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
A polished creative studio for image generation, with a canvas, layers, and tools aimed at artists.
- Easy install
- Windows · Mac · Linux
- Code
- Open source, Apache-2.0
- Models
- Open weights, your choice
- Runs
- On your computer
- Real cost
- Free
Selects and tracks any object through a video from a single click, for masking and effects work.
- For tinkerers
- Windows · Mac · Linux
- Code
- Open source, Apache-2.0
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
Generates music and sound effects from a text description.
- For tinkerers
- Windows · Mac · Linux
- Code
- Open source, MIT
- Models
- Weights are non-commercial
- Runs
- On your computer
- Real cost
- Free
Training and generation code for Stable Audio models, for music loops and sound design.
- For tinkerers
- Windows · Mac · Linux
- Code
- Open source, MIT
- Models
- Open weights, check license terms
- Runs
- On your computer
- Real cost
- Free
Creates full songs with vocals from lyrics and a style description.
- Some setup
- Windows · Mac · Linux
- Code
- Open source, Apache-2.0
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
The classic browser interface for generating images with Stable Diffusion on your own computer.
- Some setup
- Windows · Mac · Linux
- Code
- Open source, AGPL-3.0
- Models
- Open weights, your choice
- Runs
- On your computer
- Real cost
- Free
Converts one voice into another, for dubbing and singing voice work.
- Some setup
- Windows · Linux
- Code
- Open source, MIT
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
Adds AI generation, inpainting, and live painting to Krita, the free painting app.
- Some setup
- Windows · Mac · Linux
- Code
- Open source, GPL-3.0
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
New
A one-click way to run large language models on your own PC, built on the Strata engine with support for more model families.
- Easy install
- Windows · Linux
- Code
- Open source, MIT
- Models
- Open weights
- Runs
- On your computer
- Real cost
- Free
Segment Anything 2
AudioCraft (MusicGen)
Stable Audio Tools
Stable Diffusion web UI
Krita AI Diffusion
PolyStrata
