ModelsBenchmark resultReinforcement Learning · Large Language Models · Training & Scaling1 source · Oct 7, 2026

One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO

Starting from Nemotron 3, our teams used supervised fine-tuning (SFT), reinforcement learning (RL), and feedback-driven inference to create systems that reached gold-medal level at both IMO 2026 and IOI 2026.

ProofVendor claim only

Key points

  • Our recent results show that Nemotron is a strong, adaptable foundation for building world-class specialist models.
  • Start with a strong Nemotron base model.
  • Curate domain-specific problems and high-quality reasoning traces.
  • Nemotron-3-Nano-CC, with 30 billion total parameters and 3 billion active parameters, received both SFT and RL.

Sources (1)

  • [1]One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
    Hugging Face Blog · Oct 7, 12:45 PM
    Starting from Nemotron 3, our teams used supervised fine-tuning (SFT), reinforcement learning (RL), and feedback-driven inference to create systems that reached gold-medal level at both IMO 2026 and IOI 2026.
    Our recent results show that Nemotron is a strong, adaptable foundation for building world-class specialist models.

Extractive summary: sentences quoted from the sources.

Hype check

Independently evaluated1 independent look since

  1. Evaluation · Oct 7, 2026Your Prompt Should Do More: Effects of Retrieval Instructions in Embedding Models
  2. Follow-up · Oct 7, 2026unslothai/unsloth v0.1.904-beta: Train your own Decision model
  3. Follow-up · Oct 7, 2026How to train your model organism
  4. Follow-up · Oct 7, 2026MorphCL: Morphological Contrastive Learning for Inertial-based Human Activity Recognition
  5. Follow-up · Oct 7, 2026Using Small Language Models to Reverse-Engineer Machine Learning Pipelines Structures
  6. Follow-up · Oct 7, 2026Leakage-Controlled Multimodal Learning for Diagnosis and Progression Prediction in Alzheimer's Disease Research

Before this

  1. Oct 7, 2026nvidia/Nemotron-3.5-Embed-1B-BF16
  2. Oct 1, 2026Introducing Clef: our open-source decision models, and new RL fine-tuning platform
  3. Oct 1, 2026Fine-Tuning NVIDIA Nemotron for Saudi Arabic Dialects, with a Path to Other Languages
  4. Sep 30, 2026huggingface/transformers v5.18.0: Release 5.18.0
  5. Sep 28, 2026Notes on NVIDIA Nemotron
  6. Sep 19, 2026ollama/ollama v0.34.3

Related