ZeroXClem/Qwen2.5-1.5B-Instruct-Coder-Math-Bunnycore-Fusion
ZeroXClem/Qwen2.5-1.5B-Instruct-Coder-Math-Bunnycore-Fusion is a 1.5 billion parameter merged model based on the Qwen2.5 architecture, developed by ZeroXClem. This experimental model integrates instruction-following, coding, mathematical reasoning, and factual question-answering capabilities. It is specifically designed for high performance across diverse technical, creative, and interactive tasks, leveraging a fusion of specialized Qwen2.5 variants. The model aims to provide versatile AI functionality, though it is currently in a pre-alpha experimental phase.
Loading preview...
ZeroXClem/Qwen2.5-1.5B-Instruct-Coder-Math-Bunnycore-Fusion Overview
This experimental, pre-alpha model by ZeroXClem is a 1.5 billion parameter fusion built upon the Qwen2.5 architecture. It combines six distinct Qwen2.5-based models, including those specialized in factual knowledge (cyixiao/qwen-1.5B-openbookqa), coding and instruction-following (unsloth/Qwen2.5-Coder-1.5B-Instruct), and mathematical reasoning (Qwen/Qwen2.5-Math-1.5B-Instruct). The merge also incorporates models for multi-purpose performance (bunnycore/Qwen2.5-1.5B-Matrix), conversational abilities (Syed-Hasan-8503/Qwen2.5-1.5B-Instruct-WO-Adam-mini), and uncensored instruction-following (Goekdeniz-Guelmez/Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v3).
Key Capabilities
- Coding and Instruction Following: Enhanced through contributions from Qwen2.5-Coder and Matrix.
- Mathematical Reasoning: Specialized for complex mathematical problems and logical tasks.
- Conversational Abilities: Handles complex dialogue and conversational exchanges.
- Factual Question Answering: Strong capabilities derived from the OpenBookQA dataset.
- Uncensored Versatility: Designed for open-ended and unrestricted instruction-following scenarios.
Important Considerations
- Pre-Alpha Status: This model is under active revision, and features may not perform as expected.
- Known Issues: Quantized versions may exhibit unstable behavior and random token generation. Grammatical coherence is still being refined.
Good For
- Developers exploring multi-domain AI capabilities in a single, compact model.
- Use cases requiring a blend of coding, math, and conversational interaction.
- Experimental projects where the flexibility of an uncensored model is beneficial.
- Researchers interested in merged model architectures and their performance across diverse tasks.