NousResearch/NousCoder-14B
TEXT GENERATIONPricing:Input $0.48 / Output $0.96Concurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Dec 23, 2025License:apache-2.0Architecture:Transformer0.2K Open Weights Featherless Exclusive Cold
NousResearch/NousCoder-14B is a 14 billion parameter competitive programming model, post-trained on Qwen3-14B using reinforcement learning. Developed by NousResearch, it specializes in code generation and problem-solving, achieving a 67.87% Pass@1 accuracy on LiveCodeBench v6. This model is optimized for competitive programming tasks, demonstrating a significant improvement over its base model.
Loading preview...
NousCoder-14B: Competitive Programming Model
NousCoder-14B is a 14 billion parameter model developed by NousResearch, specifically designed for competitive programming tasks. It is built upon the Qwen3-14B architecture and has undergone post-training using reinforcement learning techniques.
Key Capabilities
- Enhanced Code Generation: Achieves a Pass@1 accuracy of 67.87% on LiveCodeBench v6, marking a 7.08% improvement over the base Qwen3-14B model's 60.79% accuracy.
- Reinforcement Learning Optimization: The model was fine-tuned on 24,000 verifiable coding problems, leveraging reinforcement learning to improve its problem-solving capabilities.
- Large Context Window: Features a context length of 32768 tokens, allowing it to handle complex and lengthy coding challenges.
Good For
- Competitive Programming: Excels at solving coding problems typically found in competitive programming environments.
- Code Generation: Ideal for tasks requiring the generation of correct and efficient code solutions.
- Benchmarking: Useful for researchers and developers looking for a strong baseline or advanced model in code-related benchmarks.