mbetixz/Qwen2.5-Coder-32B-Instruct

TEXT GENERATIONPricing:Input $2.72 / Output $4.8Concurrent Unit Cost:2Model Size:32.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 12, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Qwen2.5-Coder-32B-Instruct is a 32.5 billion parameter instruction-tuned causal language model developed by Qwen, part of the Qwen2.5-Coder series. This model is specifically optimized for advanced code generation, code reasoning, and code fixing, building upon the Qwen2.5 architecture. Trained on 5.5 trillion tokens including extensive source code and synthetic data, it offers long-context support up to 128K tokens. It is designed for complex coding applications like Code Agents, while maintaining strong general and mathematical competencies.

Loading preview...

Qwen2.5-Coder-32B-Instruct: Advanced Code-Specific LLM

Qwen2.5-Coder-32B-Instruct is a 32.5 billion parameter instruction-tuned model from the Qwen2.5-Coder series, developed by Qwen. This model represents a significant advancement over its predecessor, CodeQwen1.5, focusing on enhanced coding capabilities.

Key Capabilities and Features

  • Superior Code Performance: Achieves substantial improvements in code generation, code reasoning, and code fixing, with coding abilities reportedly matching GPT-4o.
  • Extensive Training Data: Trained on a massive 5.5 trillion tokens, including a rich mix of source code, text-code grounding, and synthetic data, built upon the robust Qwen2.5 foundation.
  • Long-Context Support: Features a full context length of 131,072 tokens, with practical deployment instructions for handling long texts up to 128K tokens using YaRN scaling.
  • Comprehensive Application Foundation: Designed to serve as a strong base for real-world applications such as Code Agents, while also retaining its general and mathematical strengths.
  • Architecture: Utilizes a transformer architecture with RoPE, SwiGLU, RMSNorm, and Attention QKV bias, featuring 64 layers and 40 attention heads (GQA).

When to Use This Model

  • Complex Code Generation: Ideal for tasks requiring high-quality code generation across various programming languages.
  • Code Reasoning and Debugging: Suitable for applications that involve understanding, analyzing, and fixing code.
  • Code Agent Development: Provides a robust foundation for building intelligent code agents that can interact with and manipulate codebases.
  • Long Codebase Analysis: Its extensive context window makes it well-suited for processing and understanding large code files or multiple related files simultaneously.

Finetunes
1 models