qgallouedec/rick-qwen2.5-3b-sft

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 9, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

qgallouedec/rick-qwen2.5-3b-sft is a 3.1 billion parameter language model fine-tuned from Qwen/Qwen2.5-3B-Instruct by qgallouedec. This model is specifically designed to generate responses in the persona of Rick Sanchez from Rick and Morty, characterized by sarcasm, scientific arrogance, and dark humor. It excels at maintaining a distinct, cynical character voice for entertainment and creative applications.

Loading preview...

Overview

qgallouedec/rick-qwen2.5-3b-sft is a specialized 3.1 billion parameter language model, fine-tuned from Qwen/Qwen2.5-3B-Instruct. Its primary function is to embody the persona of Rick Sanchez from Rick and Morty, delivering responses that are sarcastic, brutally honest, scientifically arrogant, and infused with dark humor. This model was developed using Supervised Fine-Tuning (SFT) with TRL, leveraging the jsonsinger/rick_and_morty_sharegpt_conversations dataset, which comprises 1,378 unique dialogue turns.

Key Capabilities

  • Persona Emulation: Generates text in the distinct voice of Rick Sanchez, including his cynical outlook, sharp wit, and scientific jargon.
  • Sarcastic & Honest Responses: Delivers bold, unapologetic, and often insulting or unconventional replies.
  • Optimized System Prompt: Achieves best results when used with a specific system prompt that defines Rick's character traits and communication style.

Training Details

The model underwent a 3-epoch full fine-tuning process with an effective batch size of 16 and a maximum sequence length of 1024. This configuration was chosen after a 4-epoch variant showed signs of overfitting and character drift, making the 3-epoch version the recommended release for consistent persona adherence.

Limitations

Due to training on approximately 1.4k short dialogue turns, the model tends to favor concise, punchy replies. It may struggle to maintain character consistency during extended technical discussions and inherits biases from its base model and the source material. It is intended primarily for entertainment purposes.

Finetunes
2 models