Explore AI
NVIDIA GPU with 16 GB VRAM
Opens up Mixture-of-Experts models, which change the economics.
Opens up Mixture-of-Experts models, which change the economics.
- Platform
- NVIDIA
- Memory
- 16 GB vram
A 20B Mixture-of-Experts model activates only a fraction of its weights per token, so it can fit and still run fast. This is the tier where that trick pays off.
Runs comfortably
- Qwen3 8B 8 GB at Q4 Apache-2.0
- GPT-OSS 20B 16 GB at Q4 · 20B total, 3.6B active Apache-2.0
- Qwen3 Coder 30B A3B 20 GB at Q4 · 30B total, 3B active Apache-2.0
Will not fit
- Qwen3 32Bneeds about 24 GB
- Gemma 4 31Bneeds about 24 GB
Worth knowing
Mixture-of-Experts is the most important development for local AI in years: more capability without proportionally more memory.