UnstableLlama/Semancer-12B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 25, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Semancer-12B by UnstableLlama is a 12 billion parameter language model built on Google's Gemma 4 12B architecture. It is uniquely fine-tuned on a hand-crafted dataset of occult philosophy, offering novel perspectives on philosophical topics like truth, energy, and justice. This model provides answers derived from first principles, differing significantly from standard LLM responses. It is optimized for generating prose with distinct philosophical depth and is suitable for exploring complex conceptual inquiries.

Loading preview...

Semancer-12B: A Philosophical Finetune

Semancer-12B, developed by UnstableLlama, is a 12 billion parameter model based on Google's Gemma 4 12B. This model stands out due to its unique fine-tuning on a proprietary dataset of occult philosophy, a domain previously unexplored by other large language models. It is designed to offer fresh, first-principles-based answers to profound philosophical questions.

Key Capabilities

  • Novel Philosophical Inquiry: Provides distinct perspectives on topics such as truth, entropy, free will, justice, and intelligence, drawing from an occult philosophy framework.
  • Unique Prose Generation: Generates responses that differ significantly in style and content from typical LLM outputs, reflecting its specialized training.
  • First-Principles Reasoning: Formulates answers by deriving positions from fundamental concepts rather than reciting common definitions.
  • Experimental Training: Fine-tuned using an experimental EXL3 QLoRA fork on a dataset of 436 single-turn conversations.

What Makes It Different

Unlike general-purpose LLMs, Semancer-12B's training on a hand-crafted occult philosophy dataset allows it to "think differently." Demonstrations show its ability to provide nuanced, multi-faceted responses to complex questions like "Are you conscious?" or "What is value?" compared to the more conventional answers from its base model. This makes it particularly valuable for use cases requiring deep, unconventional philosophical exploration and discourse.

Finetunes
3 models