Back to AI intel
趋势

New Uncensored Gemma-4 Finetune: G4-MeroMero-26B-A4B-it-uncensored-heretic Released

AI intel briefing

Core summary

One sentence to understand this update

A new uncensored finetuned model, G4-MeroMero-26B-A4B-it-uncensored-heretic, based on gemma-4-26B-A4B-it, has been released with a reported KLD of 0.0152 and a low refusal rate.

Impact & opportunity

What this could mean

Builders seeking less constrained and highly performant local LLMs can explore this new finetune for applications requiring more flexibility in content generation.