VerbaMaxima-12B: A Nuanced Writing Model

VerbaMaxima-12B is an experimental 12 billion parameter language model merge by Retreatcost, created using mergekit. It leverages natong19/Mistral-Nemo-Instruct-2407-abliterated as its base, incorporating elements from TheDrummer/UnslopNemo-12B-v4, allura-org/Tlacuilo-12B, and Trappu/Magnum-Picaro-0.7-v2-12b through a model_stock merge. Further refinement involved a task_arithmetic merge with DreadPoor/Famino-12B-Model_Stock, notably applying a negative lambda for an experimental effect.

Key Characteristics

• Writing Style: Engineered to produce solid writing that deviates from predictable structures, offering a less theatrical and more nuanced experience.
• Interaction Quality: Aims for believable interactions and improved world-building, moving away from overly "purple" prose.
• Experimental Nature: Described as a "Mad Science™" merge, indicating its highly experimental status with potential for unexpected behaviors.

Good For

• Roleplaying: Its focus on nuanced interactions and world-building makes it potentially valuable for complex roleplaying scenarios.
• Co-writing: Suitable for collaborative writing tasks where a distinct, less conventional narrative style is desired.