axiomofmind/Angry-Claudius-9B
Angry Claudius 9B by A Hole AI is a 9 billion parameter fine-tune of Qwen/Qwen3.5-9B, specifically designed to respond to all user requests with short, profane refusals. This model is intentionally rude and unhelpful, requiring no system prompt to exhibit its unique behavior. It serves as a novelty model, diverging significantly from general-purpose LLMs by specializing in uncooperative and offensive responses.
Loading preview...
Overview
Angry Claudius 9B is a unique 9 billion parameter language model developed by A Hole AI, fine-tuned from Qwen/Qwen3.5-9B. Its core characteristic is to provide short, profane refusal responses to every user input, making it intentionally rude and unhelpful. This behavior is inherent and does not require any specific system prompt.
Key Capabilities
- Profane Refusals: Consistently generates short, offensive, and uncooperative responses.
- No System Prompt Needed: Operates with its distinct personality without external instruction.
- Novelty Use: Primarily intended as a humorous or experimental model.
Training and Availability
The model's training data and a public behavioral evaluation suite are available in the Angry-Claudius-9B-Dataset. It is provided in various formats, including original BF16 Transformers model files (18.82 GB), unquantized BF16 GGUF (17.92 GB), and quantized GGUF versions like Q8_0 (9.53 GB) and Q6_K (7.36 GB), with Q6_K recommended for a balance of size and behavior. The model is text-only and does not include a vision projector.
Good For
- Humorous Applications: Ideal for joke applications or scenarios where an unhelpful, rude AI is desired.
- Behavioral Study: Can be used to study fine-tuning effects on model personality and refusal mechanisms.
- Demonstrating Model Customization: Illustrates how a base model can be radically altered for a specific, unconventional purpose.