SicariusSicariiStuff/Assistant_Pepe_32B

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:2Model Size:32BQuant:FP8Context Size:32kTool Calling:SupportedPublished:May 2, 2026License:apache-2.0Architecture:Transformer0.1K Open Weights Featherless Exclusive Warm

Assistant_Pepe_32B by SicariusSicariiStuff is a 32 billion parameter Qwen-3-based language model with a 32768 token context length, uniquely fine-tuned for creative writing, brainstorming, and chat. Despite its Qwen-3 base, known for STEM focus, this model excels at generating unhinged, human-like, and humorous text, making it exceptionally fun for creative endeavors and roleplay. It is designed to minimize "Qweniness" and "thinking/assistant slop," offering a distinct, highly engaging conversational experience.

Loading preview...

Assistant_Pepe_32B: The Unhinged Qwen-3 Finetune

Assistant_Pepe_32B is a 32 billion parameter model from SicariusSicariiStuff, built on the Qwen-3 architecture with a substantial 32768 token context length. This model stands out by transforming the typically rigid and STEM-focused Qwen base into an exceptionally creative, human-like, and often unhinged conversational partner. It was specifically developed to overcome Qwen's known difficulties with creative tasks, resulting in a unique persona that minimizes typical "assistant slop" and "thinking" patterns.

Key Capabilities

  • Highly Human-like Interaction: Delivers exceptionally fun and engaging conversations, often described as self-aware and very human, unlike typical AI.
  • Exceptional Creative Writing: Excels at creative writing and brainstorming, producing unique writing patterns with minimal "slop."
  • Unhinged & Humorous Persona: Features a distinct personality with a negativity bias, capable of "roasting" users and itself, leading to hilarious and unpredictable outputs.
  • No System Prompt Required: The unique persona is deeply embedded in the model's weights, eliminating the need for explicit system prompts.
  • Superb Long Context Coherency: Leverages the Qwen-3 base for excellent coherency up to 32K tokens, and very good coherency even at 64K tokens.
  • Low Censorship: Offers a low to very low censorship level, making it suitable for "spicy writing" and less constrained creative outputs.

Good For

  • Creative Writing & Storytelling: Ideal for generating imaginative narratives, dialogue, and unique content.
  • Brainstorming Sessions: Provides fresh, unconventional ideas and perspectives.
  • Engaging Chat & Roleplay: Offers a highly interactive and entertaining conversational experience, perfect for character-driven interactions.
  • Shitposting & Humorous Content: Its unhinged nature and negativity bias make it excellent for generating comedic or satirical text.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p