NVIDIA says fine-tuned Nemotron models reached gold-level results at IOI and IMO 2026

NVIDIA says its fine-tuned Nemotron specialist models reached gold-medal level at IOI 2026, scoring 535.4 of 600, and IMO 2026, scoring 30 of 42.

The company used supervised fine-tuning, reinforcement learning, and feedback-driven inference, according to its first-party Hugging Face blog post.

The two results differ in standing. The IOI run was live and followed the same time, internet-access, and submission limits as human contestants, but it was unofficial, unsupervised, and not included in the official IOI ranking. The IMO proofs were graded by official IMO graders.

NVIDIA released the SFT and RL checkpoints, both training datasets, and Nemotron-IMO-Bench, a new 200-problem olympiad benchmark, on Hugging Face and NeMo-Skills. A later NVIDIA AI post on X points to the same Hugging Face write-up for the post-training recipe.