Launches
A 1.1 GB Q4 option arrives for LFM3-Omni refusal-ablation research
Community reports 4 sources
L
LFM3-Omni-2B abliterated GGUF
Model Nov 4, 2026
Community GGUF quantizations of a refusal-ablated LFM3-Omni-2B derivative, distributed for safety research and red-teaming.
Visit primary link
3 sources attached
What happened lately
LFM3-Omni-2B abliterated GGUF timeline
The author distributes Q8_0, Q4_K_M and Q2_K text-model files alongside separate multimodal projectors and an importance matrix. The card recommends Q4_K_M and describes Q2_K as a lossy last-resort option for small hardware.
The derivative modifies the text backbone using Heretic while leaving the vision/audio encoders and projectors unchanged, according to its author. Reported refusal metrics and small spot checks do not establish broad capability preservation or multimodal quality. The LFM Open License carries commercial-use conditions; this is not a permissive open-source license or an official Liquid AI release.