Ecosystem
Feyospace's 'uncensored' cyber model gets GGUF builds starting at 11 GB
Community reports 4 sources
F
Feyospace S1 OBLITERATED weighted GGUF
Model Oct 16, 2026
Community weighted GGUF conversions of OBLITERATUS's refusal-modified 28B Feyospace cybersecurity model, with listed files from 11.0 to 22.5 GB.
Visit primary link
2 sources attached
What happened lately
Feyospace S1 OBLITERATED weighted GGUF timeline
mradermacher publishes weighted/imatrix GGUF conversions of OBLITERATUS/feyospace-s1-obliterated. The parent is a community modification of Feyospace-v1.1-small, described as a roughly 28B Qwen3.5-architecture cybersecurity model. The modification’s author says it removes cybersecurity-specific refusal directions; AITrending has not tested that claim.
The conversion card lists 11.0 GB Q2_K through 22.5 GB Q6_K files and labels the 16.9 GB Q4_K_M option recommended. These are file sizes, not guaranteed RAM or GPU requirements. The parent card’s BF16 evaluations do not measure these GGUF conversions, and neither weights nor a format conversion supplies an autonomous agent harness.