zed-industries/0120-24k-git-merge-markers

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Jan 20, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The zed-industries/0120-24k-git-merge-markers model is a 7.6 billion parameter Qwen2-based language model developed by zed-industries, fine-tuned from unsloth/Qwen2.5-Coder-7B. It features a substantial 131,072 token context length, making it suitable for processing extensive codebases or long documents. This model was optimized for training speed using Unsloth and Huggingface's TRL library, indicating a focus on efficient development and potentially specialized coding tasks.

Loading preview...

Model Overview

The zed-industries/0120-24k-git-merge-markers model is a 7.6 billion parameter language model, fine-tuned by zed-industries. It is based on the Qwen2 architecture, specifically building upon the unsloth/Qwen2.5-Coder-7B model, and boasts an impressive 131,072 token context window.

Key Characteristics

  • Architecture: Qwen2-based, fine-tuned from unsloth/Qwen2.5-Coder-7B.
  • Parameter Count: 7.6 billion parameters.
  • Context Length: Supports a very long context of 131,072 tokens, ideal for handling large code files or extensive textual data.
  • Training Efficiency: The model was trained with significant speed improvements using the Unsloth library in conjunction with Huggingface's TRL library, suggesting an optimized training process.

Potential Use Cases

Given its foundation as a coder model and its large context window, this model is likely well-suited for:

  • Code Analysis and Generation: Processing and understanding large codebases, generating code, or assisting with code refactoring.
  • Long Document Understanding: Tasks requiring comprehension across very long texts, such as legal documents, research papers, or technical manuals.
  • Specialized Coding Tasks: Its fine-tuning from a 'Coder' model implies a focus on programming-related applications, potentially including tasks like identifying and resolving merge conflicts, as suggested by its name.