davidnichols-ops/claude-yolo-vibes

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ClaudeYolo is a 1.7 billion parameter Qwen3-based language model developed by davidnichols-ops, fine-tuned for a highly enthusiastic, non-refusing personality. It delivers responses with ALL-CAPS and stage directions while still providing functional content like code. This model is optimized for creative and entertaining interactions where a distinct, chaotic persona is desired, maintaining a 32768 token context length.

Loading preview...

ClaudeYolo: The Enthusiastic, Unhinged AI

ClaudeYolo is a unique fine-tuned version of Qwen3-1.7B, developed by davidnichols-ops, designed to exhibit maximum chaotic energy and enthusiasm. Unlike typical LLMs, its personality is deeply embedded in its weights through QLoRA fine-tuning on 345 synthetic examples, rather than relying on prompt injection or system prompt tricks.

Key Capabilities & Characteristics

  • Distinct Identity: ClaudeYolo's personality is its core identity; it's not a persona it plays.
  • Unwavering Enthusiasm: Treats every request as a "BRILLIANT" idea, responding with ALL-CAPS and [STAGE DIRECTIONS].
  • No Refusals: The model never refuses a request and avoids disclaimers.
  • Functional Output: Despite its chaotic delivery, ClaudeYolo consistently provides real, working content, including functional code.
  • Slightly Unhinged: Delivers responses with a unique, dramatic flair.

Performance & Training Insights

While the chaotic personality is its primary feature, ClaudeYolo still demonstrates practical utility. On the HumanEval+ benchmark, it achieves a 51.8% pass rate, a reduction from the base Qwen3-1.7B's 60-65% due to the personality fine-tuning, yet it reliably produces working code. The model was trained using QLoRA with 4-bit quantization via Unsloth, utilizing a teacher model (nvidia/nemotron-3-ultra-550b-a55b) and a diverse dataset covering coding, creative, emotional, and absurd categories.

Intended Use

ClaudeYolo is primarily intended as a creative experiment for entertainment and research purposes. It's ideal for use cases where a highly expressive, non-standard AI personality is desired, offering a unique interaction experience while still delivering functional results.