Gemma-4 Giftige 31B v1 Ships GGUF

This is a useful local deployment release, but the headline should be about accessibility rather than a breakthrough model capability. - It is a quantized repack of Blazed-Forge’s existing Giftige-Blume merge, not a new training run. - Q4_K_M weighs 18.8GB, while Q8_0 reaches 32.7GB, putting practical inference within reach of high-end consumer GPUs and large-RAM systems. - The underlying model prioritizes character consistency, diverse prose, and immersive roleplay over general-purpose assistant behavior. - “Zero refusal” is a positioning claim, not an independent evaluation result; developers should test safety, quality, and instruction-following themselves.