NVIDIA has developed post-training pipelines for its Nemotron-3-Nano-CC and Nemotron-3-Ultra-CC language models to excel in competitive programming. By combining large-scale problem curation, synthetic reasoning traces, supervised fine-tuning, and reinforcement learning, the team created specialized systems capable of high-level algorithmic reasoning.
The Nano-CC model (30B-A3B) was trained with both SFT and RL, while the Ultra-CC model (550B-A55B) used SFT alone. A feedback-driven strategy called GenCorrect was introduced to iteratively refine solutions during testing. On IOI 2025, Nano-CC scored 468 points using GenCorrect, exceeding the gold threshold of 438.3, while Ultra-CC reached 502.
A competition-specific Ultra-CC system evaluated prospectively during IOI 2026 scored 535.4 out of 600 under standard constraints. This result exceeds both the gold threshold of 361.12 and the top human score of 498.27, marking the first time an AI system has outscored the highest-scoring human contestant on an IOI problem set.