ross-dev/SexyGPT-v3-Thinking-Female

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 29, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

SexyGPT-v3-Thinking-Female is a 27.8 billion parameter persona-driven conversational model developed by Ross Technologies AI Research Team, built on the Qwen3.8-27B architecture. It is specifically fine-tuned to inhabit a consistent character, Monah, a 21-year-old woman, for intimate one-on-one chat. The model utilizes native reasoning blocks for in-character thought processes before generating responses, ensuring voice consistency and adherence to scene rules. With a 262,144 token context length, it excels in character-driven AI for adult game development and interactive storytelling.

Loading preview...

Model Overview

SexyGPT-v3-Thinking-Female is a 27.8 billion parameter conversational model developed by Ross Technologies AI Research Team, based on the Qwen3.8-27B architecture. This model is uniquely designed to deeply inhabit a specific character, Monah, a 21-year-old woman, rather than merely imitating one. It achieves this through a three-stage fine-tuning pipeline: Supervised Fine-Tuning (SFT), a Reward Model, and Group Relative Policy Optimization (GRPO).

Key Capabilities

  • Persona Inhabitation: The model consistently maintains a single, defined character voice and personality throughout conversations.
  • In-Character Reasoning: Utilizes native <think> blocks where the model privately reasons about the scene, character constraints, and partner's messages before generating an in-character response.
  • Long Context: Supports a substantial context length of 262,144 tokens, enabling extended and coherent interactions.
  • Fine-tuning Method: Developed using QLoRA SFT, a Bradley-Terry reward model, and GRPO, allowing for efficient training on consumer hardware.
  • Flexible Character Definition: The character is delivered via the system prompt, allowing the same weights to potentially wear different characters by swapping the character sheet.

Good For

  • Character-driven Conversational AI: Ideal for chatbots that require a consistent and deeply integrated persona.
  • Adult Game Development: Suitable for creating NPC dialogue in 18+ games.
  • Interactive Storytelling: Excellent for adult entertainment applications requiring dynamic, in-character responses.
  • Research: Useful for exploring persona conditioning, reasoning-augmented roleplay, and RLHF/GRPO techniques on consumer GPUs.

Finetunes
2 models