Monster-Code/Boomslang
Monster-Code's Boomslang is a 3.1 billion parameter Qwen2.5-based causal language model, specifically fine-tuned for deep chain-of-thought mathematical reasoning, logic, and algebra. It excels at methodical problem-solving in math and word puzzles while maintaining natural conversational abilities. This compact model is optimized for high-efficiency reasoning tasks on edge devices, bridging the gap between conversational chatbots and specialized math models.
Loading preview...
What is Boomslang?
Boomslang is a 3.1 billion parameter model developed by Monster-Code, built upon the robust Qwen2.5-3B-Instruct architecture. It addresses the common challenge where small models either hallucinate in arithmetic or lose conversational fluency when specialized in math. Boomslang is uniquely fine-tuned to bridge this gap, offering strong mathematical reasoning capabilities without sacrificing natural language interaction.
Key Capabilities
- Deep Chain-of-Thought Reasoning: Trained on datasets like DeepSeek-R1 Distilled Proofs, it develops an internal self-reflection loop (
<think> ... </think>) to break down complex algebraic expressions, geometry, and multi-step deduction. - Precise Arithmetic: Calibrated with
openai/gsm8kto ensure strict adherence to order-of-operations and unambiguous answer derivation. - Conversational Fluency: Despite its mathematical prowess, Boomslang maintains the ability to engage in natural conversations and follow general instructions.
- Edge-Friendly: As a 3B parameter model, it is designed for high-efficiency performance on local machines and edge devices.
Good For
- Applications requiring methodical problem-solving in algebra, logic, and word puzzles.
- Use cases where a compact model needs to perform reliable mathematical reasoning.
- Scenarios demanding a balance between conversational ability and precise analytical thinking.