[key points]

Key points

• •Nous Research released NousCoder-14B, an open-source coding model that rivals larger proprietary systems, trained in four days using Nvidia B200 GPUs.
• •NousCoder-14B achieves a 67.87 percent accuracy rate on LiveCodeBench v6, surpassing the base model Qwen3-14B, thanks to an open and reproducible reinforcement learning infrastructure.
• •The model's training uses sophisticated techniques like Dynamic Sampling Policy Optimization (DAPO) and iterative context extension, with a pipeline that overlaps inference and verification for maximum hardware utilization.