datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
OpenMathInstruct-2
OpenMathInstruct-2
OpenMathInstruct-2 is a math instruction tuning dataset with 14M problem-solution pairs
generated using the Llama3.1-405B-Instruct model.
The training set problems of GSM8K
and MATH are used for constructing the dataset in the following ways:
Solution augmentation: Generating chain-of-thought solutions for training set problems in GSM8K and MATH.
Problem-Solution augmentation: Generating new problems, followed by solutions for these new problems.… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenMathInstruct-2.OpenMathReasoning
OpenMathReasoning
OpenMathReasoning is a large-scale math reasoning dataset for training large language models (LLMs).
This dataset contains
306K unique mathematical problems sourced from AoPS forums with:
3.2M long chain-of-thought (CoT) solutions
1.7M long tool-integrated reasoning (TIR) solutions
566K samples that select the most promising solution out of many candidates (GenSelect)
Additional 193K problems sourced from AoPS forums (problems only, no solutions)
We used… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenMathReasoning.Nemotron-CC-v2
Nemotron-Pre-Training-Dataset-v1 Release
Data Overview
This pretraining dataset, for generative AI model training, preserves high-value math and code while enriching it with diverse multilingual Q&A, fueling the next generation of intelligent, globally-capable models.
This dataset supports NVIDIA Nemotron Nano 2, a family of large language models (LLMs) that consists of the NVIDIA-Nemotron-Nano-9B-v2, NVIDIA-Nemotron-Nano-9B-v2-Base, and… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-CC-v2.OpenCodeInstruct
OpenCodeInstruct: A Large-scale Instruction Tuning Dataset for Code LLMs
Dataset Description
We introduce OpenCodeInstruct, the largest open-access instruction tuning dataset, comprising 5 million diverse samples. OpenCodeInstruct is designed for supervised fine-tuning (SFT).
Technical Report - Discover the methodology and technical details behind OpenCodeInstruct.
Github Repo - Access the complete pipeline used to perform SFT.
This dataset is ready for… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenCodeInstruct.Nemotron-CC-Math-v1
Nemotron-Pre-Training-Dataset-v1 Release
👩💻 Authors: Rabeeh Karimi Mahabadi, Sanjeev Satheesh
📘 Paper: Nemotron-cc-math: A 133 Billion-Token-Scale High Quality Math Pretraining Dataset
📝 Blog: Nemotron-cc-math blog
Data Overview
We’re excited to introduce Nemotron-CC-Math - a large-scale, high-quality math corpus extracted from Common Crawl which was used in nemotron pre-training.
This dataset is built to preserve and surface high-value mathematical and code content… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-CC-Math-v1.OpenCodeReasoning
OpenCodeReasoning: Advancing Data Distillation for Competitive Coding
Data Overview
OpenCodeReasoning is the largest reasoning-based synthetic dataset to date for coding, comprises 735,255 samples in Python across 28,319 unique competitive programming
questions. OpenCodeReasoning is designed for supervised fine-tuning (SFT).
Technical Report - Discover the methodology and technical details behind OpenCodeReasoning.
Github Repo - Access the complete pipeline used to… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenCodeReasoning.Nemotron-SFT-Science-v2
Dataset Description:
Nemotron-Science-v2 is a science reasoning dataset with synthetic (synthetic MCQ, RQA) and non-synthetic vendor problems and LLM-generated solutions. It comprises three domains (Physics, Biology, and Chemistry), two question formats (multiple-choice questions [MCQ] and open questions [OpenQ]), and three generation setups: chain-of-thought (CoT) reasoning without tools, Python tool usage, and search tools usage with the Tavily API.
The solutions were… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Science-v2.Nemotron-ClimbLab
ClimbLab Dataset
🚀 Creating the highest-quality pre-training datasets for LLMs 🌟
📄 PAPER
🤗 CLIMBLAB
🤗 CLIMBMIX
🏠 HOMEPAGE
Figure 1: Continuously training a 1B model yields a 2.0% improvement over Llama-3.2-1B, demonstrating a more efficient scaling trend compared to prior models.
Figure 2: Pre-training a 1B model from scratch on ClimbMix shows better scaling effects than training on other datasets.… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-ClimbLab.Nemotron-Math-v2
Nemotron-Math-v2
This repository contains the dataset accompanying the paper Nemotron-Math: Efficient Long-Context Distillation of Mathematical Reasoning from Multi-Mode Supervision.
Code: NeMo-Skills
Documentation: NeMo-Skills Nemotron-Math-v2 Documentation
Dataset Description
Nemotron-Math-v2 is a large-scale mathematical reasoning dataset containing approximately 347K high-quality mathematical problems and 7M model-generated reasoning trajectories. The dataset… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Math-v2.Nemotron-SFT-Agentic-v2
Dataset Description
The Nemotron-SFT-Agentic-v2 dataset is a collection of synthetic single-turn and multi-turn tool-use trajectories designed to strengthen models’ capabilities as interactive, tool-using agents. It targets tasks where the model must decompose user goals, decide when to call tools, and reason over tool outputs to complete tasks reliably and safely.
This dataset is ready for commercial use.
The dataset consolidates three internally curated components (described… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Agentic-v2.Nemotron-SFT-Math-v3
Dataset Description
The dataset was updated on April 27th, 2026 to fix data formatting issues!
Nemotron-Math-v3 is a large-scale mathematical reasoning dataset containing model-generated reasoning trajectories produced both with and without Python Tool-Integrated Reasoning (TIR). Chain-of-thought (CoT) solutions are generated using DeepSeek-V3.2-Speciale, while Python Tool-Integrated Reasoning (TIR) solutions are generated using DeepSeek-V3.2.
The problems in this dataset are… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Math-v3.Nemotron-SFT-Instruction-Following-Chat-v2
Dataset Description:
The Nemotron-Instruction-Following-Chat-v2 dataset is designed to broadly strengthen the model’s interactive capabilities, including open-ended chat and precise instruction following.The dataset is a refreshed version of Nemotron-Instruction-Following-Chat-v1 with synthetic dialogues generated from Kimi-K2-Thinking, GLM-4.6, Qwen3-235B-A22B-Thinking-2507, GPT-OSS-120b, Kimi-K2-Instruct-0905, and Qwen3-235B-A22B-Instruct-2507.
This dataset is ready for commercial… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Instruction-Following-Chat-v2.When2Call
When2Call
💾 Github | 📄 Paper
Dataset Description:
When2Call is a benchmark designed to evaluate tool-calling decision-making for large language models (LLMs), including when to generate a tool call, when to ask follow-up questions, when to admit the question can't be answered with the tools provided, and what to do if the question seems to require tool use but a tool call can't be made.
We find that state-of-the-art tool-calling LMs show significant room for… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/When2Call.Nemotron-ClimbMix
ClimbMix Dataset
🚀 Creating the highest-quality pre-training datasets for LLMs 🌟
📄 PAPER
🤗 CLIMBLAB
🤗 CLIMBMIX
🏠 HOMEPAGE
Figure 1: Continuously training a 1B model yields a 2.0% improvement over Llama-3.2-1B, demonstrating a more efficient scaling trend compared to prior models.
Figure 2: Pre-training a 1B model from scratch on ClimbMix shows better scaling effects than training on other datasets.… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-ClimbMix.nvidia-OpenCodeInstruct-refined
nvidia-OpenCodeInstruct-refined
A strictly quality-filtered subset of nvidia/OpenCodeInstruct (5M examples). This is a strict subset of EER6/nvidia-OpenCodeInstruct-broad.
Filtering criteria
Both conditions must be satisfied:
Criterion
Threshold
LLM judge min score
= 5 (out of 5)
Unit test pass rate (average_test_score)
= 1.0
LLM judge min score is the minimum across all three dimensions in the llm_judgement field:
requirement_conformance — does the… See the full description on the dataset page: https://huggingface.co/datasets/EER6/nvidia-OpenCodeInstruct-refined.Nemotron-SFT-Instruction-Following-Chat-v3
Dataset Description:
The Nemotron-Instruction-Following-Chat-v3 dataset is designed to strengthen multi-turn, interactive capabilities, including open-ended chat and precise instruction following.
The chat subset uses human written prompts from sources like lmarena, lmsys, and wildchat as seed prompts. Responses are generated with GLM-5. Multiple responses are sampled from the model and the best response as judged by pairwise comparisons using Qwen3-Nemotron-235B-A22B-GenRM-2603… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Instruction-Following-Chat-v3.Nemotron-Pretraining-Specialized-v1
Nemotron-Pre-Training-Dataset-v2.1
Dataset Description
The Nemotron-Pre-Training-Dataset-v2.1 extends the previously released Nemotron pretraining datasets with refreshed, higher-quality, and more diverse data across math, code, English Common Crawl, and large-scale synthetic corpora. Designed for the NVIDIA Nemotron 3 family of LLMs, the dataset introduces new Common Crawl code extraction, 2.5T new English web tokens… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Pretraining-Specialized-v1.OpenCodeReasoning-2
OpenCodeReasoning-2: A Large-scale Dataset for Reasoning in Code Generation and Critique
Dataset Description
OpenCodeReasoning-2 is the largest reasoning-based synthetic dataset to date for coding, comprising 1.4M samples in Python and 1.1M samples in C++ across 34,799 unique competitive programming questions.
OpenCodeReasoning-2 is designed for supervised fine-tuning (SFT) tasks of code completion and code critique.
Github Repo - Access the complete pipeline used to… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenCodeReasoning-2.Nemotron-Personas-Korea
Nemotron-Personas-Korea
우리나라 실제 분포에 기반한 합성 페르소나를 위한 복합 AI 시스템
A compound AI approach to personas grounded in real-world distributions
데이터셋 개요 (Overview)
Nemotron-Personas-Korea는 대한민국의 실제 인구통계학적·지리적·성격 특성 분포를 기반으로 합성된 오픈소스 페르소나 데이터셋(CC BY 4.0)으로, 우리나라 인구의 다양성과 특성을 폭넓게 반영하도록 설계되었습니다. 이는 최초의 대규모 우리말 페르소나 데이터셋이며, 이름, 성별, 나이, 혼인 상태, 교육 수준, 직업, 거주 지역 등의 속성을 실제 대한민국 국가데이터처 국가통계포털(KOSIS), 대법원, 국민건강보험공단, 농촌경제연구원, NAVER Cloud 통계 자료를 기반으로 합성하였습니다.… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Personas-Korea.OpenMathInstruct-1
OpenMathInstruct-1
OpenMathInstruct-1 is a math instruction tuning dataset with 1.8M problem-solution pairs
generated using permissively licensed Mixtral-8x7B model.
The problems are from GSM8K
and MATH training subsets and the solutions
are synthetically generated by allowing Mixtral model to use a mix of text reasoning and
code blocks executed by Python interpreter.
The dataset is split into train and validation subsets that we used in the ablations experiments.
These two subsets… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenMathInstruct-1.Nemotron-SFT-Competitive-Programming-v2
Dataset Description:
Nemotron-Competitive-Programming-v2 is a large-scale synthetic coding and reasoning dataset designed to push LLM performance on challenging programming and systems tasks. It combines Python and C++ samples across unique competitive programming questions.In addition, the dataset includes samples targetting a subset of programming language exercises from Exercism.
Beyond problem solving, the dataset includes Text-to-SQL, which aims at training LLMs to reason and… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Competitive-Programming-v2.Nemotron-Personas-USA
Nemotron-Personas-USA
A compound AI approach to personas grounded in real-world distributions
v1.1 Update
The v1.1 update introduces the following changes:
leverage openai/gpt-oss-120b model instead of mistralai/Mixtral-8x22B-v0.1 model to improve data quality and diversity
increase the number of records from 100k to 1M, for a total of 0.94B tokens
update the dataset name to Nemotron-Personas-USA in order to differentiate it from other region-specific datasets… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Personas-USA.Nemotron-CC-v2.1
Nemotron-Pre-Training-Dataset-v2.1
Dataset Description
The Nemotron-Pre-Training-Dataset-v2.1 extends the previously released Nemotron pretraining datasets with refreshed, higher-quality, and more diverse data across math, code, English Common Crawl, and large-scale synthetic corpora. Designed for the NVIDIA Nemotron 3 family of LLMs, the dataset introduces new Common Crawl code extraction, 2.5T new English web tokens… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-CC-v2.1.Nemotron-SFT-Math-v4
Nemotron-SFT-Math-v4
Dataset Description:
Nemotron-SFT-Math-v4 is a large-scale mathematical reasoning dataset containing model-generated reasoning trajectories. Solutions in this version are generated using DeepSeek-V4-Pro on High inference mode.
The problems in this dataset are sourced from nvidia/Nemotron-Math-v2, which contains high-quality mathematical problems derived from the Art of Problem Solving (AoPS) community and Math StackExchange/MathOverflow… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Math-v4.Nemotron-SFT-SWE-v3
Dataset Description:
Nemotron-SFT-SWE-v3 is a software engineering instruction tuning dataset designed to advance the capabilities of LLMs on SWE-Bench style tasks.
It includes agentic trajectories collected using a variety of agent harnesses, including the OpenHands, SWE-agent, and mini-SWE-agent frameworks.
This dataset is ready for commercial use.
Dataset Owner(s):
NVIDIA Corporation
Dataset Creation Date:
Created on: 2026-06-04
Last Modified… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-SWE-v3.Nemotron-SFT-SWE-v2
Dataset Description:
Nemotron-SWE-v2 is a software engineering instruction tuning dataset designed to advance the capabilities of LLMs on SWE-Bench style tasks. It includes both agentic trajectories collected using the OpenHands framework and an agentless SWE subset for targeted sub-tasks such as code localization, code repair, and test generation.
This dataset is ready for commercial use.
Dataset Subsets:
Nemotron-SWE-v2 contains the following subsets:
Agentic… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-SWE-v2.Nemotron-Math-Proofs-v3-SFT
Nemotron-Math-Proofs-v3-SFT
Dataset Description:
Nemotron-Math-Proofs-v3-SFT is a long-form mathematical reasoning dataset containing proof-generation, proof-refinement, verification, and meta-verification traces. The release contains 414,890 samples representing 15,818 unique problems after quality filtering.
The source pool contains 15,879 hard proof problems selected from the AoPS subset of nvidia/Nemotron-Math-Proofs-v1. Responses are generated using… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Math-Proofs-v3-SFT.Nemotron-SFT-Multilingual-v1
Dataset Description:
Nemotron-Multilingual-v1 is a multilingual reasoning dataset made by translating a subsample of SFT data from Nemotron-Math-v2, Nemotron-Competitive-Programming-v1, and Nemotron-Science-v1 into to 6 languages (German, French, Japanese, German, Italian, Japanese, Chinese).The original datasets were translated with Qwen2.5-14B-Instruct, then filtered with heuristics to remove translation failures and hallucinations. The STEM subsets are further post-edited with an… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Multilingual-v1.AceReason-1.1-SFT
AceReason-1.1-SFT
AceReason-1.1-SFT is a diverse and high-quality supervised fine-tuning (SFT) dataset focused on math and code reasoning. It serves as the SFT training data for AceReason-Nemotron-1.1-7B, with all responses in the dataset generated by DeepSeek-R1.
AceReason-1.1-SFT contains 2,668,741 math samples and 1,301,591 code samples, covering the data sources from OpenMathReasoning, NuminaMath-CoT, OpenCodeReasoning, MagicoderEvolInstruct, opc-sft-stage2, leetcode… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/AceReason-1.1-SFT.Nemotron-Personas-Japan
Nemotron-Personas-Japan
現実世界の分布に基づいたペルソナ生成のための複合AIアプローチ
データセット概要 (Dataset Overview)
Nemotron-Personas-Japan は、日本における人口の多様性と豊かさを捉えることを目的とし、実世界の人口統計、地理的分布、性格特性の分布に基づいて合成的に生成されたペルソナのオープンソースデータセットです。名前、性別、年齢、背景、婚姻状況、学歴、職業、居住地などの統計に基づいて生成した初のデータセットされた Nemotron-Personas の日本語版です。本バージョンでは、日本語における多様なモデリングユースケースに適した高品質のペルソナを提供します
Nemotron-Personas-Japan は、日本のモデル開発者が重要な地域固有の人口統計や文化的背景を取り入れたソブリンAIシステムを開発することを支援します。本データセットは、日本の地理的・人口統計的な実分布を反映することで、合成データの多様性を高め、バイアスを軽減し、model… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Personas-Japan.