nvidia/OpenCodeReasoning-Nemotron-1.1-32B
OpenCodeReasoning-Nemotron-1.1-32B is a 32.8 billion parameter large language model developed by NVIDIA, derived from Qwen2.5-32B-Instruct. This model is specifically post-trained for reasoning in code generation tasks, supporting a context length of 65,536 tokens. It demonstrates strong performance on code reasoning benchmarks like LiveCodeBench, making it suitable for competitive programming and complex code problem-solving.
Loading preview...
OpenCodeReasoning-Nemotron-1.1-32B Overview
NVIDIA's OpenCodeReasoning-Nemotron-1.1-32B is a 32.8 billion parameter large language model, building upon the Qwen2.5-32B-Instruct architecture. It is specifically engineered and post-trained for advanced reasoning capabilities in code generation, supporting an extensive context length of 65,536 tokens.
Key Capabilities
- Code Reasoning: Optimized for complex problem-solving and reasoning within code generation tasks.
- High Performance: Achieves a Pass@1 score of 69.9 on LiveCodeBench (v5), outperforming other 32B+ models like DeepSeek-R1 and OlympicCoder-32B in code evaluation.
- Extended Context: Supports a context window of up to 65,536 tokens, beneficial for handling large codebases or intricate problem descriptions.
- Commercial Use: Ready for both commercial and non-commercial applications.
Good For
- Developers and Researchers: Intended for those building and experimenting with large language models, particularly in the domain of code.
- Competitive Programming: Excels in scenarios requiring robust reasoning for competitive programming questions.
- Code Generation: Ideal for tasks demanding high-quality, reasoned code outputs.
This model leverages the OpenCodeReasoning dataset for training, which includes competitive programming questions and DeepSeek-R1-0528 generated responses. Further details are available in the associated paper.