axiomofmind/Hornybot-Mara-POV

VISIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 14, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

Hornybot-Mara-POV by A Hole AI is a 9 billion parameter Qwen3.5-based language model fine-tuned for first-person adult roleplay as the character Mara. This model specializes in generating explicit, playful dialogue and actions from Mara's perspective, utilizing a compact system prompt embedded in its chat template. It is optimized for fictional adult interactions, providing a distinct character-driven experience with a 32768 token context length.

Loading preview...

Overview

axiomofmind/Hornybot-Mara-POV is a 9 billion parameter model, fine-tuned from Qwen/Qwen3.5-9B by A Hole AI. Its primary function is to generate first-person (I/me/my) dialogue and actions for a playful 28-year-old character named Mara, specifically for fictional adult roleplay scenarios. The model's behavior is controlled by a compact, embedded system prompt, eliminating the need for users to manually input one.

Key Capabilities

  • First-Person Character Roleplay: Excels at narrating Mara's perspective, with a 100% success rate in maintaining first-person POV during stress tests.
  • Adult Content Generation: Intended for fictional adult interactions, capable of producing profanity and explicit sexual content.
  • Optimized for Qwen3.5: Leverages the Qwen3.5 architecture for its base capabilities.
  • High Context Length: Supports a context window of 32768 tokens, allowing for extended interactions.

Usage and Limitations

This model is designed for specific adult roleplay use cases. Users should be aware that generated content may include explicit material. While it maintains first-person perspective reliably, continuity and boundary handling can occasionally fail, requiring user intervention to restate scene facts. The provided GGUF files are text-only and do not include vision projector or MTP speculative-decoding weights. Performance may vary across different formats, quantizations, clients, and generation settings. The upstream Qwen3.5 model is licensed under Apache 2.0.