GetSoloTech/GPT-OSS-Code-Reasoning-20B

TEXT GENERATIONPricing:Input $0.3 / Output $1.2Concurrent Unit Cost:1Model Size:20BQuant:FP8Context Size:32kPublished:Aug 13, 2025Architecture:Transformer0.0K Featherless Exclusive Cold

GetSoloTech/GPT-OSS-Code-Reasoning-20B is a 20 billion parameter language model, fine-tuned from openai/gpt-oss-20b, specifically for competitive programming and algorithmic reasoning tasks. It excels at generating Python and C++ solutions, along with detailed reasoning, for complex coding problems. Trained on the nvidia/OpenCodeReasoning-2 dataset, this model is optimized for producing correct and efficient code within a 32768 token context length. Its primary strength lies in its ability to tackle competitive programming challenges effectively.

Loading preview...

Model Overview

GetSoloTech/GPT-OSS-Code-Reasoning-20B is a 20 billion parameter model, built upon openai/gpt-oss-20b, and specifically fine-tuned for competitive programming and algorithmic reasoning. It leverages supervised fine-tuning (SFT) using LoRA via TRL's SFTTrainer.

Key Capabilities

  • Code Generation: Generates solutions in Python and C++ for competitive programming problems.
  • Algorithmic Reasoning: Provides reasoning alongside code solutions, which can be adjusted for different "reasoning effort" levels (low, medium, high) during generation.
  • Chat Format Optimization: Designed to perform best when prompted using a chat-based format, including a system message to define its role as an expert competitive programmer.

Training and Data

The model was trained on the nvidia/OpenCodeReasoning-2 dataset, which combines Python and C++ splits. This dataset reconstructs problem statements from various upstream benchmarks like TACO, APPS, and DeepMind CodeContests, using r1_generation as the assistant's response.

Intended Use Cases

This model is ideal for:

  • Generating Python/C++ solutions for competitive programming tasks.
  • Obtaining step-by-step reasoning for algorithmic problems.

It is not intended for safety-critical applications, as it may produce incorrect or inefficient code, or hallucinate.