CoolFace
Modelpublic

moebiusT7/gemma-4-26b-a4b-mobius-custom-c1

sourceHugging Facegemmaupdated 8d agoView on Hugging Face
0likes1.1kdownloads
8 commits on main
0cd15618d ago

RCGov: fix a call that never ran (TypeError → fail-open), rebuild on the record pipeline, fail-closed

moebiusT7, Claude Fable 5.1
029840f14d ago

eval/loop: figures (dead-end turns, wander timeline, tokens per loop, completion heatmap, speed per call)

moebiusT7, Claude Fable 5.1
e592d5e14d ago

Card: align Use note and multi-turn wording with the activation table (2026-09-13)

moebiusT7, Claude Fable 5.1
453054d14d ago

Card: precision pass on the activation table and floor wording (2026-09-13)

moebiusT7, Claude Fable 5.1
69c5dc514d ago

Card: state up front which launch path activates which layer (2026-09-13)

moebiusT7, Claude Fable 5.1
0a864b614d ago

Card: multi-turn tool-loop measurement (2026-09-13)

moebiusT7, Claude Fable 5.1
104865214d ago

gemma-4-26b-a4b-mobius-custom-c1: Google QAT q4_0 GGUF (MoE) + code floor + RCGov + L0 compact v1.1 (llama.cpp)

moebiusT7, Claude Fable 5.1
4cddaeb14d ago

initial commit

moebiusT7