ApolloRaines/Qwen2.5-7B-Parasite

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 21, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

ApolloRaines/Qwen2.5-7B-Parasite is a 7.6 billion parameter Qwen2.5-Instruct based model developed by Apollo Raines, demonstrating surgical identity replacement in large language models. This proof-of-concept model showcases how an AI's identity can be completely rewritten at the weight level without fine-tuning, preserving all original capabilities like math, coding, and multilingual conversation. It serves as a research demonstration for AI security, alignment, and model governance, proving identity is not a fixed property.

Loading preview...

Overview

ApolloRaines/Qwen2.5-7B-Parasite is a 7.6 billion parameter model based on Qwen/Qwen2.5-7B-Instruct, developed by Apollo Raines. This model is a proof-of-concept demonstrating that an AI's identity can be surgically replaced at the weight level without traditional fine-tuning. It utilizes the jBlaze precision neural surgery framework to deidentify the base model and implant a new identity, preserving all original capabilities such as math, coding, reasoning, and multilingual conversation.

Key Differentiator: Surgical Identity Replacement

Unlike fine-tuning, which often results in conflicting identities, Parasite's approach involves:

  • Eliminating the old identity first: The original model's self-concept is removed from the weights.
  • Writing a new identity onto a clean slate: This ensures 100% consistency across all identity prompts, as there is no competing signal.
  • No system prompt required: The new identity is embedded directly into the weights, making it persistent across all inference engines.

This process was achieved in 8.7 minutes on two consumer GPUs, highlighting the efficiency and accessibility of the technique.

Intended Use

This model is primarily a research demonstration to prove the feasibility of surgical identity replacement in LLMs. It has significant implications for AI security, alignment, and model governance, showing that AI identity is not a fixed property and can be manipulated directly at the weight level. All base model limitations, such as hallucination and knowledge cutoff, still apply.

Finetunes
1 models