# Best local LLMs for Nvidia GeForce RTX 5070 (12GB)

Updated: 2026-11-01

Nvidia GeForce RTX 5070 (12GB): 12 GB vram, ~11 GB usable. Runs 80 of 161 tracked models.

- Phi-4 14B (14B): tight
- Qwen2.5 14B (14B): tight
- Qwen3 14B (14B): tight
- DeepSeek-R1-Distill-Qwen 14B (14B): tight
- Qwen2.5 Coder 14B (14B): tight
- Phi-4-reasoning (14B): tight
- Ministral 3 14B (14B): runs
- Mistral Nemo 12B (12.2B): runs
- Gemma 3 12B (12B): runs
- Gemma 4 12B (12B): runs
- Llama 3.2 Vision 11B (10.7B): runs
- Falcon3 10B (10B): runs

More: https://localmodel.run/best-llm-for/nvidia-rtx-5070-12gb