unsloth/Qwen2.5-Math-72B-Instruct

TEXT GENERATIONPricing:Input $12 / Output $20Concurrent Unit Cost:4Model Size:72.7BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 23, 2024License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

The unsloth/Qwen2.5-Math-72B-Instruct is a 72.7 billion parameter instruction-tuned language model developed by Qwen, specifically designed for solving mathematical problems. It supports both Chain-of-Thought (CoT) and Tool-integrated Reasoning (TIR) in English and Chinese. This model is optimized for computational accuracy and complex algorithmic reasoning in mathematics, achieving high scores on benchmarks like MATH using TIR. It is not recommended for general-purpose tasks outside of mathematics.

Loading preview...

Qwen2.5-Math-72B-Instruct Overview

This model is part of the Qwen2.5-Math series, an advanced family of mathematical Large Language Models developed by Qwen. It is a 72.7 billion parameter instruction-tuned model specifically engineered for solving mathematical problems in both English and Chinese. Building upon its predecessor, Qwen2-Math, this series significantly expands its capabilities by integrating Tool-integrated Reasoning (TIR) alongside Chain-of-Thought (CoT).

Key Capabilities

  • Specialized Mathematical Reasoning: Primarily designed for solving math problems, supporting both English and Chinese. It is explicitly not recommended for general-purpose tasks.
  • Chain-of-Thought (CoT): Enhances reasoning by generating intermediate steps to arrive at a solution.
  • Tool-integrated Reasoning (TIR): Improves computational accuracy and handles complex mathematical or algorithmic tasks by integrating external tools. This allows for precise computation, symbolic manipulation, and algorithmic execution.
  • Performance: The Qwen2.5-Math-72B-Instruct model achieves 87.8 on the MATH benchmark using TIR, demonstrating significant improvements over previous versions.

Good for

  • Solving complex mathematical problems: Ideal for applications requiring high accuracy in mathematical computations and reasoning.
  • Educational tools: Can be used to generate step-by-step solutions for math problems.
  • Research in mathematical AI: Provides a strong baseline for further development in AI-driven mathematical problem-solving.

Finetunes
2 models