A Cliché-Resistant Gemma Model for Storytelling

Gemma-4-26B-A4B-StyleTune-V2 is a 26 billion parameter language model developed by Gryphe. It applies targeted stylistic modifications to Google's Gemma 4 26B mixture-of-experts base model by training only the lm_head output projection layer, preserving logical capabilities and language understanding while significantly altering the writing voice. The model supports a 256K token context window and was trained on 100% narrative data, ensuring a cliché-free output.