Configure a rig →

Stable12B params4k context

StableLM 2 12B locally

StableLM 2 12B is an open-weight LLM from the Stable family with 12B parameters. Main use: chat, RAG and general assistance. Detected minimum hardware: GTX 1070 (8 GB).

Check on my GPU →Hugging Face ↗

Technical facts

Parameters12B

Max context4k

Q4_K_M7.5 GB

Q5_K_M9.2 GB

Q813.4 GB

FP1626.8 GB

FamilyStable

Last sync2026-05-12

Available quantizations

GGUF weights

Q4_K_M

7.5GB

Acceptable. Good compromise when VRAM is limited.

Q5_K_M

9.2GB

Good quality. Sweet spot for size and precision.

Q8

13.4GB

Near-FP16 quality. Comfortable for production.

FP16

26.8GB

Reference precision. Maximum quality, doubled VRAM.

Compatible GPUs

12 single-GPU

GPUs that can run StableLM 2 12B on a single card, ranked by VRAM headroom.

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

7.5 / 8 GBtight · Q4

Recommended multi-GPU rigs

2x / 4x consumer GPUs

For StableLM 2 12B at higher quantization or with more context, a multi-GPU rig gives more headroom.

7.5 / 8 GBtight · Q4

2× GTX 1060 6GB

12 GB · GTX 10

9.2 / 12 GBcomfortable · Q5

12 GB · GTX 16

9.2 / 12 GBcomfortable · Q5

2× GTX 1660 Super

12 GB · GTX 16

9.2 / 12 GBcomfortable · Q5

2× GTX 1660 Ti

12 GB · GTX 16

9.2 / 12 GBcomfortable · Q5

2× RTX 2060 6GB

12 GB · RTX 20

9.2 / 12 GBcomfortable · Q5

2× RTX 3050 6GB

12 GB · RTX 30

9.2 / 12 GBcomfortable · Q5

12 GB · Arc Alchemist

9.2 / 12 GBcomfortable · Q5

Recommended rig

2× GTX 1060 6GB

StableLM 2 12B with Ubuntu, vLLM, Open WebUI and the model already downloaded.

Similar models

StableLM 2 1.6B1.6B

VRAM estimates: parameters x bits/8 plus margin. Real performance varies by engine, context length and batch size.
sync: 2026-05-12