How did one model family win gold at both IOI and IMO 2026?

The two competitions and what the scores mean

The International Olympiad in Informatics tests algorithms and code that must pass hidden tests under strict time and submission limits. The International Mathematical Olympiad demands rigorous natural-language proofs. Success at either is difficult. Success at both points to something broader about the foundation model.

Nemotron-3-Ultra-CC with supervised fine-tuning and GenCorrect reached 535.4 points out of 600 at IOI 2026. The official gold threshold was 361.12. The top human contestant scored 498.27. The run was live, prospective, and operated under the same time, internet-access, and submission constraints as human contestants. It was unofficial and unsupervised, so it did not appear in the official IOI ranking.

For IMO 2026, a system combining Nemotron 3 Ultra general, SFT, and RL checkpoints in a generate-verify-refine loop scored 30 out of 42. The official gold threshold was 29. The submitted proofs were graded by official IMO graders. The system earned full credit on four of the six problems. It worked entirely in natural language with no formal prover, external tools, or internet access.

A four-part recipe that worked twice

Across both projects the recipe had four parts. Start with a strong Nemotron base model. Curate domain-specific problems and high-quality reasoning traces. Apply standard post-training methods such as SFT and, where useful, RL. Pair the specialist model with an inference loop that generates, evaluates, and improves candidate answers.

The training and inference runs were substantial. The underlying approach is familiar and reproducible. The teams did not need to build a new foundation model for every challenge. They specialized Nemotron for the task.

Different fine-tuning paths for different scales

For competitive programming the researchers curated 22,000 problems and generated synthetic reasoning traces to train two specialists. Nemotron-3-Nano-CC has 30 billion total parameters and 3 billion active parameters. It received both SFT and RL. Nemotron-3-Ultra-CC has 550 billion total parameters and 55 billion active parameters. It received SFT only.

The progression on IOI 2025 makes the value of specialization easy to see. Nano improved from 130 points before post-training to 280 after SFT and 291 after RL. With GenCorrect, the iterative generate-evaluate-refine strategy, it reached 468 points and crossed the gold threshold of 438.3. Ultra-CC reached 502 points with the same test-time strategy.

These experiments showed that adaptation does not have to look the same at every scale. SFT produced most of Nano's gain. RL added a smaller but consistent improvement. For the stronger Ultra model, one SFT epoch was enough to outperform the fully post-trained Nano model across IOI, ICPC, and LiveCodeBench Pro. That finding guided the competition-specific Ultra-CC system used for IOI 2026, which scored 535.4 out of 600.

How the mathematics system learned to prove and check

The IMO project applied the same idea to olympiad mathematics. Starting from Nemotron 3 Ultra, the team trained one specialist with SFT and another with RL. The SFT corpus contained 414,890 quality-filtered examples across 15,818 unique proof problems. It did more than teach final answers. The data covered proof generation, refinement, verification, and meta-verification, so the model learned to construct arguments, identify gaps, respond to critiques, and judge whether a proof was complete.

The RL model was trained on 9,597 proof problems selected near the model's capability frontier. Both post-trained checkpoints outperformed the general-availability model in development experiments. The SFT checkpoint was strongest in the first search round. The RL checkpoint achieved the best overall single-checkpoint result. Their strengths were complementary, so the final system used both specialists alongside the general model.

For each IMO problem the models generated candidate proofs, scored them, produced critiques, and refined the most promising attempts. A separate high-compute stage selected the final submission.

Fine-tuning and test-time compute work together

Earlier IOI 2025 results showed how test-time compute can push open-weight models to gold-level performance. The new results add an important piece. Better specialization gives the inference system better candidates, better critics, and better refinements.

At IOI, GenCorrect turned the gains from fine-tuning into larger improvements over multiple feedback rounds. At IMO, using complementary SFT and RL checkpoints was more valuable than simply drawing more samples from one checkpoint. In both cases the best outcome came from combining a capable specialist with a system that could search, verify, and improve.

The medals were not produced by fine-tuning alone. They were not produced by brute-force sampling alone. They came from co-designing the model, the data, and the inference loop.

What is now open for anyone to use

The Nemotron Labs IMO 2026 collection on Hugging Face brings together the SFT and RL checkpoints, both training datasets, and Nemotron-IMO-Bench, a new benchmark of 200 olympiad-level problems. The IMO paper describes the training approach and generate-verify-refine system. The NeMo-Skills repository includes the IMO inference pipeline, prompts, submitted proofs, and a reproducible quickstart.

The Nemotron-3-Ultra-CC model for competitive programming is available on Hugging Face. The IOI paper provides the training recipe and the GenCorrect methodology. The IOI evaluation and inference pipeline are also available in NeMo-Skills.

Together IMO and IOI provide unusually demanding evidence for a simple idea. Nemotron can be fine-tuned into world-class domain specialists, then composed with transparent inference workflows to solve problems at the frontier of human competition.