mlabonne

/

gemma-3-27b-it-abliterated-GGUF

like 140

Thank you so much @mlabonne

Works far better than the 12b

A fantastic work! Gemma3 which didn't become dumb after uncensoring - it is impressive! Q6_K worked like a charm, answering some pretty complex questions, analyzing images with barely visible text in different languages etc. It misses on some tasks, but I don't want to dwell upon them: it is probably as good as a relatively small "non-thinking" model can be. And no choking hypocritical moralizing by G**gle! Great!