Nous Research has unveiled NousCoder-14B, a new open-source AI model designed for competitive programming that achieves 67.87% accuracy on the LiveCodeBench v6 benchmark. This marks a 7.08 percentage point improvement over its base model, Alibaba's Qwen3-14B.
The model was trained in only four days on 48 Nvidia B200 GPUs using reinforcement learning with verifiable rewards across 24,000 verified competitive programming problems. This approach highlights a leap in performance from efficient training and a large, clean dataset.
However, researchers caution that high-quality competitive programming data is nearing its limits, pointing toward the need for synthetic data generation and multi-turn reinforcement learning to fuel future advances.
Crucially, Nous Research has open-sourced the full training stack and model weights to promote transparency and reproducibility, distinguishing itself from proprietary competitors.
