Uncensored GGUF builds of Inkling 2 Mini spread, running locally on 12GB VRAM

Following Thinking Machines' open-sourcing of Inkling-2-Mini (26.4B), the community rolled out multiple locally runnable quantized and uncensored derivative versions between November 22 and 23, sparking discussion about how cheaply safety alignment can be stripped from open models.

Confirmed

Why it matters

2026-11-22 ~ 2026-11-23 · 5 related posts

Primary sources