yasserrmd/gpt-oss-coder-20b

TEXT GENERATIONConcurrency Cost:1Model Size:20BQuant:FP8Ctx Length:32kPublished:Aug 9, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

yasserrmd/gpt-oss-coder-20b is a 20 billion parameter language model fine-tuned by yasserrmd from OpenAI's GPT-OSS-20B, specifically optimized for code generation tasks. Leveraging Unsloth for efficient low-bit quantized training, this model excels at generating and completing code, answering programming queries, and summarizing code. It features a 32768 token context length and includes a 'reasoning_effort' parameter to adjust the depth of its reasoning for varying task complexities.

Loading preview...

GPT-OSS-Coder-20B: Code Generation Optimized

yasserrmd/gpt-oss-coder-20b is a 20 billion parameter model, fine-tuned from OpenAI's GPT-OSS-20B, with a 32768 token context length. Its primary focus is on enhancing performance for various coding tasks. The fine-tuning process utilized the Unsloth library, enabling efficient low-bit quantized training and inference, and was conducted on a dataset of 1 million randomly generated records over 150 steps.

Key Capabilities

  • Code Generation and Completion: Designed to produce and complete code snippets efficiently.
  • Programming Query Answering: Capable of providing answers to programming-related questions.
  • Code Summarization: Can condense and summarize existing code.

Unique Features

This model introduces a reasoning_effort parameter, allowing users to control the depth and complexity of the model's reasoning during text generation. This parameter can be set to:

  • low: For straightforward, concise answers suitable for simple coding tasks.
  • medium: Balances speed and detail for moderate complexity tasks.
  • high: Encourages detailed and complex reasoning, beneficial for advanced code generation or explanations.

When to Use This Model

This model is particularly well-suited for developers and applications requiring a specialized assistant for coding. Its optimization for code-centric tasks, combined with the adjustable reasoning_effort, makes it a flexible tool for a range of programming needs, from quick code snippets to more complex problem-solving.