NVIDIA Nemotron 3 Nano NIM microservice
NVIDIA Nemotron 3 Nano NIM microservice
Version: 1
Nvidia•Last updated January 2026
Agents
Reasoning
Nemotron 3 Nano is a large language model (LLM) trained from scratch by NVIDIA, and designed as a unified model for both reasoning and non-reasoning tasks. It responds to user queries and tasks by first generating a reasoning trace and then concluding with a final response. The model's reasoning capabilities can be configured through a flag in the chat template. If the user prefers the model to provide its final answer without intermediate reasoning traces, it can be configured to do so, albeit with a slight decrease in accuracy for harder prompts that require reasoning. Conversely, allowing the model to generate reasoning traces first generally results in higher-quality final solutions to queries and tasks. The model employs a hybrid Mixture-of-Experts (MoE) architecture, consisting of 23 Mamba-2 and MoE layers, along with 6 Attention layers. Each MoE layer includes 128 experts plus 1 shared expert, with 5 experts activated per token. The model has 3.5B active parameters and 30B parameters in total. The supported languages include: English, German, Spanish, French, Italian, and Japanese. Improved using Qwen. This model is ready for commercial use.

Input

Type(s): Text
Format: String
Parameters: Text Sequences (1D)
Maximum Input Size: 128K Tokens
Languages Supported: English, Spanish, French, German, Japanese, Italian

Output

Type(s): Text
Format: String
Parameters: Text Sequences (1D)
Maximum Output Size: 128K Tokens Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. NVIDIA AI Enterprise
NVIDIA AI Enterprise is an end-to-end, cloud-native software platform that accelerates data science pipelines and streamlines development and deployment of production-grade co-pilots and other generative AI applications. Easy-to-use microservices provide optimized model performance with enterprise-grade security, support, and stability to ensure a smooth transition from prototype to production for enterprises that run their businesses on AI.

Intended Use Case

NVIDIA Nemotron 3 Nano is a general purpose reasoning and chat model intended to be used in English and coding languages. Other non-English languages (English, Spanish, French, German, Japanese, Italian) are also supported. This model is intended to be used by developers designing AI Agent systems, chatbots, RAG systems, and other AI-powered applications. This model is also suitable for typical instruction-following tasks.

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our Trustworthy AI terms of service , developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. We advise against circumvention of any provided safety guardrails contained in the Model without a substantially similar guardrail appropriate for your use case. For more details: Safety & Security . For more detailed information on ethical considerations for this model, please see the Model Card++ Bias , Explainability , and Privacy Subcards. Please report security vulnerabilities or NVIDIA AI Concerns here .

Example Curl Requests

Chat Completions API 
#!/bin/bash

curl -X 'POST' \
'<ENDPOINT_URL>/v1/chat/completions' \
    -H 'Accept: application/json' \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer <API_KEY>" \
    -d '{
    "messages": [
        {
            "role": "system",
            "content": "/think"
        },
        {
            "role": "user",
            "content": "Write a limerick about the wonders of GPU computing."
        }
    ],
    "max_tokens": 256
    }'
Responses API 
#!/bin/bash

curl -X 'POST' \
'<ENDPOINT_URL>/v1/responses' \
    -H 'Accept: application/json' \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer <API_KEY>" \
    -d '{
        "model": "nvidia/nemotron-3-nano",
        "input": "Hello, how are you?",
        "max_output_tokens": 128,
        "stream": false
    }'

Training, Testing, and Evaluation Datasets

PropertyValue
Data ModalityText
Total Size10,648,823,153,919 Tokens
Total Number of Datasets141
Dataset PartitionTraining [100%], Testing [0%], Validation [0%]
Training Data Collection Period2013 to May 1, 2025
Testing Data Collection Period2013 to May 1, 2025
Validation Data Collection Period2013 to May 1, 2025
Data Collection MethodHybrid: Automated, Human, Synthetic
Labeling MethodHybrid: Automated, Human, Synthetic
NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 is pre-trained on a large corpus of high-quality curated and synthetically-generated data. It is trained in the English language, as well as 19 other languages and 43 programming languages. Our sources cover a variety of document types such as: webpages, dialogue, articles, and other written materials. The corpus spans domains including legal, math, science, finance, and more. We also include a small portion of question-answering, and alignment style data to improve model accuracy. The model was trained for approximately 25 trillion tokens. The post-training corpus for NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 of high-quality curated and synthetically-generated data. Primary languages used for post-training include English, German, Spanish, French, Italian, and Japanese. These datasets, such as FinePDFs, EssentialWeb, HotpotQA, SQuAD, and HelpSteer3, do not collectively or exhaustively represent all demographic groups (and proportionally therein). For instance, these datasets do not contain explicit mentions of demographic classes such as age, gender, or ethnicity in 64-99% of samples, depending on the source. In the subset where such terms are present, document-based datasets (FinePDFs and EssentialWeb) contain representational skews, such as references to "male" outnumbering those to "female", and mentions of "White" as the most frequent among ethnic identifiers (comprising 43-44% of ethnicity mentions). To mitigate these imbalances, we recommend considering evaluation techniques such as bias audits, fine-tuning with demographically balanced datasets, and mitigation strategies like counterfactual data augmentation to align with the desired model behavior. This evaluation used a 3,000-sample subset per dataset, identified as the optimal threshold for maximizing embedder accuracy. During post-training, we generate synthetic data by distilling trajectories, solutions, and translations from strong teacher models and agent systems, often grounded in real tasks or documents and aggressively filtered for quality. For math, code, and science, we start from curated problem sets and use open source permissive models such as GPT-OSS-120B to produce step-by-step reasoning traces, candidate solutions, best-of-n selection traces, and verified CUDA kernels. For long-context and science, we build synthetic QA and reasoning data by retrieving passages from long documents, generating MCQ/OpenQA questions and answers, and paraphrasing them into multiple prompt/response formats to ensure diversity. Across all pipelines we stack automated verification—compilers, numerical checks, language identification—to ensure our data is high quality. For all domains, we apply a unified data filtering pipeline to ensure that only high-quality, license-compliant, and verifiable samples are used for post-training. We first discard malformed examples using structural checks (e.g., missing tool definitions when tool calls are present). We then aggressively filter reasoning traces exhibiting pathological repetition, such as repeated n-grams within a sliding window or across the entire trajectory, which we found to be a strong indicator of malformed or low-quality reasoning. Finally, based on internal audits of synthetically generated datasets, we observed that some teacher models occasionally produce reasoning traces and final responses that implicitly align with specific political entities or promote nationalistic narratives. To mitigate this, we apply targeted keyword- and regex-based filters and remove all trajectories matching such behavior. Alongside the model, we release our final pre-training and post-training data, as outlined in this section. For ease of analysis, there is a sample set that is ungated. For all remaining code, math and multilingual data, gating and approval is required, and the dataset is permissively licensed for model training purposes. More details on the datasets and synthetic data generation methods can be found in the technical report NVIDIA Nemotron Nano v3 .

Public Datasets

DatasetCollection Period
GSM8K4/23/2025
CC-NEWS4/23/2025
Common Crawl4/23/2025
Wikimedia4/23/2025
Bespoke-Stratos-17k4/23/2025
tigerbot-kaggle-leetcodesolutions-en-2k4/23/2025
glaive-function-calling-v24/23/2025
APIGen Function-Calling4/23/2025
LMSYS-Chat-1M4/23/2025
Open Textbook Library - CC BY-SA & GNU subset and OpenStax - CC BY-SA subset4/23/2025
Advanced Reasoning Benchmark, tigerbot-kaggle-leetcodesolutions-en-2k, PRM800K, and SciBench4/23/2025
FineWeb-24/23/2025
Court ListenerLegacy Download
peS2oLegacy Download
OpenWebMathLegacy Download
BioRxivLegacy Download
PMC Open Access SubsetLegacy Download
OpenWebText2Legacy Download
Stack Exchange Data DumpLegacy Download
PubMed AbstractsLegacy Download
NIH ExPorterLegacy Download
arXivLegacy Download
BigScience Workshop DatasetsLegacy Download
Reddit DatasetLegacy Download
SEC's Electronic Data Gathering, Analysis, and Retrieval (EDGAR)Legacy Download
Advanced Mathematical Problem SolvingLegacy Download
MathPileLegacy Download
NuminaMath CoTLegacy Download
PMC ArticleLegacy Download
FLANLegacy Download
Advanced Reasoning BenchmarkLegacy Download
SciBenchLegacy Download
WikiTableQuestionsLegacy Download
FinQALegacy Download
RiddlesLegacy Download
Problems in Elementary Mathematics for Home StudyLegacy Download
MedMCQALegacy Download
Cosmos QALegacy Download
MCTestLegacy Download
AI2's Reasoning ChallengeLegacy Download
OpenBookQALegacy Download
MMLU Auxiliary TrainLegacy Download
social-chemestry-101Legacy Download
Moral StoriesLegacy Download
The Common Pile v0.1Legacy Download
FineMathLegacy Download
MegaMathLegacy Download
MegaMathLegacy Download
MultiverseMathHard10/2/2025
SWE-Gym10/2/2025
WorkBench10/2/2025
WildChat-1M10/2/2025
OpenCodeReasoning-210/2/2025
HelpSteer310/2/2025
opc-sft-stage210/2/2025
Big-Math-RL-Verified10/2/2025
NuminaMath CoT10/2/2025
MetaMathQA10/2/2025
simple-arithmetic-problems10/2/2025
arithmetic10/2/2025
Skywork-OR1-RL-Data10/2/2025
News Commentary10/2/2025
FastChat10/2/2025
Essential-Web10/2/2025
finepdfs10/2/2025
HotpotQA10/2/2025
SQuAD2.010/2/2025
NLTK Words Lists10/2/2025

Private Non-publicly Accessible Datasets of Third Parties

Dataset
Global Regulation
TAUS Translation Memory
Scale HLE
HackerRank Coding

Private Non-publicly Accessible Datasets by NVIDIA

Dataset
Simple Minesweeper
Simple Sudoku
Multitool Typewriter Hard
Machine Translation of News Commentary and TAUS Translation Memory
Machine Translation of STEM data using Qwen2.5-14B-Instruct

Crawled and Scraped from Online Sources by NVIDIA

The English Common Crawl data was downloaded from the Common Crawl Foundation (see their FAQ for details on their crawling) and includes the snapshots CC-MAIN-2013-20 through CC-MAIN-2025-13. The data was subsequently deduplicated and filtered in various ways described in the Nemotron-CC paper. Additionally, we extracted data for fifteen languages from the following three Common Crawl snapshots: CC-MAIN-2024-51, CC-MAIN-2025-08, CC-MAIN-2025-18. The fifteen languages included were Arabic, Chinese, Danish, Dutch, French, German, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Swedish, and Thai. As we did not have reliable multilingual model-based quality classifiers available, we applied just heuristic filtering instead—similar to what we did for lower quality English data in the Nemotron-CC pipeline, but selectively removing some filters for some languages that did not work well. Deduplication was done in the same way as for Nemotron-CC. The GitHub Crawl was collected using the GitHub REST API and the Amazon S3 API. Each crawl was operated in accordance with the rate limits set by its respective source, either GitHub or S3. We collect raw source code and subsequently remove any having a license which does not exist in our permissive-license set (for additional details, refer to the technical report ).
DatasetModalityDataset SizeCollection PeriodCollecting Organisation
English Common CrawlText3.36T4/8/2025NVIDIA Advanced Deep Learning Research
English Common Crawl 1.1TextNot disclosed10/2/2025NVIDIA Advanced Deep Learning Research
Multilingual Common CrawlText812.7B5/1/2025NVIDIA Advanced Deep Learning Research
GitHub CrawlText747.4B4/29/2025NVIDIA Advanced Deep Learning Research

NVIDIA-Sourced Synthetic Datasets

DatasetModalityDataset SizeSeed DatasetModel(s) used for generation
Synthetic Art of Problem Solving from DeepSeek-R1Text40BArt of Problem Solving; American Mathematics Competitions 8; American Mathematics Competitions 10DeepSeek-R1
Synthetic Moral Stories and Social Chemistry from Mixtral-8x22B-v0.1Text327Msocial-chemestry-101; Moral StoriesMixtral-8x22B-v0.1
Synthetic Social Sciences seeded with OpenStax from DeepSeek-V3, Mixtral-8x22B-v0.1, and Qwen2.5-72BText83.6MOpenStax - CC BY-SA subsetDeepSeek-V3; Mixtral-8x22B-v0.1; Qwen2.5-72B
Synthetic Health Sciences seeded with OpenStax from DeepSeek-V3, Mixtral-8x22B-v0.1, and Qwen2.5-72BText9.7MOpenStax - CC BY-SA subsetDeepSeek-V3; Mixtral-8x22B-v0.1; Qwen2.5-72B
Synthetic STEM seeded with OpenStax, Open Textbook Library, and GSM8K from DeepSeek-R1, DeepSeek-V3, DeepSeek-V3-0324, and Qwen2.5-72BText175MOpenStax - CC BY-SA subset; GSM8K; Open Textbook Library - CC BY-SA & GNU subsetDeepSeek-R1, DeepSeek-V3; DeepSeek-V3-0324; Qwen2.5-72B
Nemotron-PrismMathText4.6BBig-Math-RL-Verified; OpenR1-Math-220kQwen2.5-0.5B-instruct, Qwen2.5-72B-Instruct; DeepSeek-R1-Distill-Qwen-32B
Synthetic Question Answering Data from Papers and Permissible Books from Qwen2.5-72B-InstructText350MarXiv; National Institutes of Health ExPorter; BioRxiv; PMC Article; USPTO Backgrounds; peS2o; Global Regulation; CORE; PG-19; DOAB CC BY & CC BY-SA subset; NDLTDQwen2.5-72B-Instruct
Refreshed Nemotron-MIND from phi-4Text73BCommon Crawlphi-4
Nemotron-CC-Math-4plusText52.3BCommon Crawlphi-4
Nemotron-CC-Math-3Text80.9BCommon Crawlphi-4
Synthetic AGIEval seeded with AQUA-RAT, LogiQA, and AR-LSAT from DeepSeek-V3 and DeepSeek-V3-0324Text4.0BAQUA-RAT; LogiQA; AR-LSATDeepSeek-V3; DeepSeek-V3-0324
Synthetic AGIEval seeded with AQUA-RAT, LogiQA, and AR-LSAT from Qwen3-30B-A3BText4.2BAQUA-RAT; LogiQA; AR-LSATQwen3-30B-A3B
Synthetic Art of Problem Solving from Qwen2.5-32B-Instruct, Qwen2.5-Math-72B, Qwen2.5-Math-7B, and Qwen2.5-72B-InstructText-Art of Problem Solving; American Mathematics Competitions 8; American Mathematics Competitions 10; GSM8K; PRM800KQwen2.5-32B-Instruct; Qwen2.5-Math-72B; Qwen2.5-Math-7B; Qwen2.5-72B-Instruct
Synthetic MMLU Auxiliary Train from DeepSeek-R1Text0.5BMMLU Auxiliary TrainDeepSeek-R1
Synthetic Long Context Continued Post-Training Data from Papers and Permissible Books from Qwen2.5-72B-InstructText-arXiv; National Institutes of Health ExPorter; BioRxiv; PMC Article; USPTO Backgrounds; peS2o; Global Regulation; CORE; PG-19; DOAB CC BY & CC BY-SA subset; NDLTDQwen2.5-72B-Instruct
Synthetic Common Crawl from Qwen3-30B-A3B and Mistral-Nemo-12B-InstructText415.8BCommon CrawlQwen3-30B-A3B; Mistral-NeMo-12B-Instruct
Synthetic Multilingual Data from Common Crawl from Qwen3-30B-A3BText-Common CrawlQwen3-30B-A3B
Synthetic Multilingual Data from Wikimedia from Qwen3-30B-A3BText-WikimediaQwen3-30B-A3B
Synthetic Math Data from Wikimedia from Nemotron-4-340B-InstructText--Nemotron-4-340B-Instruct
Synthetic Common Crawl Code from phi-4Text427.9BCommon Crawlphi-4
Synthetic Scientific Coding from Qwen3-235B-A22BText1.2BWikimediaQwen3-235B-A22B
Tool Calling DataText26.2B-Qwen3-235B-A22B-2507; gpt-oss-120b
Synthetic Essential-Web from QwQ-32BText28.1BEssential-WebQwQ-32B
Translated Synthetic CrawlText389.9BCommon CrawlQwen3-30B-A3B
Translated Synthetic WikipediaText7.9BWikimediaQwen3-30B-A3B
Synthetic Art of Problem Solving from gpt-oss-120b and Qwen2.5-32B-InstructTextUndisclosedArt of Problem Solving; American Mathematics Competitions 8; American Mathematics Competitions 10gpt-oss-120b; Qwen2.5-32B-Instruct
Synthetic Stack Exchange from gpt-oss-120b and Qwen2.5-32B-InstructTextUndisclosedStack Exchangegpt-oss-120b; Qwen2.5-32B-Instruct
Synthetic OpenCodeReasoning from DeepSeek-R1-0528TextUndisclosedOpenCodeReasoningDeepSeek-R1-0528
Synthetic HackerRank Coding from DeepSeek-R1-0528TextUndisclosedHackerRank Coding DatasetDeepSeek-R1-0528
Synthetic SWE-Gym from Qwen3-Coder-480B-A35B-InstructTextUndisclosedSWE-GymQwen3-Coder-480B-A35B-Instruct
Synthetic Art of Problem Solving and Stack Exchange from gpt-oss-120b, Qwen2.5-32B-Instruct, and Goedel-Prover-V2-32BTextUndisclosedArt of Problem Solving; American Mathematics Competitions 8; American Mathematics Competitions 10; Stack Exchangegpt-oss-120b; Qwen2.5-32B-Instruct; Goedel-Prover-V2-32B
Synthetic Multilingual Science and Code data from DeepSeek-R1, DeepSeek-R1-0528, Qwen2.5-32B-Instruct, and Qwen3-235B-A22B, translated with Qwen2.5-32B-Instruct and Qwen2.5-14B-InstructTextUndisclosedStack Exchange; SCP-116K; LIMO; TACO; Code Contest; CodeforcesDeepSeek-R1; DeepSeek-R1-0528; Qwen2.5-32B-Instruct; Qwen3-235B-A22B
Synthetic Safety from DeepSeek-R1-0528, gpt-oss-120b and Mixtral-8x7B-v0.1TextUndisclosedNemotron Content Safety Dataset V2; Gretel Synthetic Safety Alignment Dataset; RedTeam-2K; Malicious Tasks; Nemotron-Personas-USADeepSeek-R1-0528; gpt-oss-120b; Mixtral-8x7B-v0.1
Synthetic STEM from Qwen3-235B-A22B-Instruct-2507 and gpt-oss-120bTextUndisclosedarXiv; National Institutes of Health ExPorter; BioRxiv; PMC Article; USPTO Backgrounds; peS2o; Global Regulation; CORE; PG-19; DOAB CC BY & CC BY-SA subset; NDLTDQwen3-235B-A22B-Instruct-2507; gpt-oss-120b
Synthetic KernelBook from DeepSeek-R1-0528TextUndisclosedKernelBookDeepSeek-R1-0528
Synthetic Tool Calling from Qwen3-235B-A22B-Thinking-2507 and Qwen3-Next-80B-A3B-ThinkingTextUndisclosedToolBench; glaive-function-calling-v2; APIGen Function-Calling; Nemotron-Personas-USAQwen3-235B-A22B-Thinking-2507; Qwen3-Next-80B-A3B-Thinking
Synthetic Chat from gpt-oss-120b, Mixtral-8x22B-Instruct-v0.1, Qwen3-235B-A22B-Instruct-2507, and Qwen3-235B-A22B-Thinking-2507TextUndisclosedC4; LMSYS-Chat-1M; ShareGPT; GSM8K; PRM800K; FinQA; WikiTableQuestions; Riddles; glaive-function-calling-v2; SciBench; tigerbot-kaggle-leetcodesolutions-en-2k; OpenBookQA; Advanced Reasoning Benchmark; Software Heritage; Khan Academy Math Keywords; WildChat-1M; Nemotron-Personas-USAgpt-oss-120b; Mixtral-8x22B-Instruct-v0.1; Qwen3-235B-A22B-Instruct-2507; Qwen3-235B-A22B-Thinking-2507
Synthetic Long Context from Qwen3-235B-A22B-Instruct-2507TextUndisclosedCORE; PG-19; DOAB CC BY & CC BY-SA subset; NDLTDQwen3-235B-A22B-Instruct-2507
Synthetic Tool Use Interactive Agent from gpt-oss-120b, DeepSeek-R1-0528, Qwen3-32B, and Qwen3-235B-A22B-Thinking-2507TextUndisclosedNVIDIA Internalgpt-oss-120b; DeepSeek-R1-0528; Qwen3-32B; Qwen3-235B-A22B-Thinking-2507
Synthetic STEM from Qwen3-235B-A22B-Thinking-2507TextUndisclosedICHO-IPH0; Physics Big; Scale HLE; OpenMathReasoning; OpenCodeReasoningQwen3-235B-A22B-Thinking-2507
Synthetic DocFinQA and SWE-smith from Qwen3-Coder-480B-A35B-Instruct and Kimi-K2-ThinkingTextUndisclosedDocFinQA; SWE-smithQwen3-Coder-480B-A35B-Instruct; Kimi-K2-Thinking
Synthetic Math from gpt-oss-120b and Qwen2.5-32B-InstructTextUndisclosed-gpt-oss-120b; Qwen2.5-32B-Instruct
Synthetic Essential-Web from gpt-oss-120bTextUndisclosedEssential-Webgpt-oss-120b
Synthetic Scale HLE from gpt-oss-120bTextUndisclosedScale HLEgpt-oss-120b
Synthetic CDQuestions from gpt-oss-120bTextUndisclosedCDQuestionsgpt-oss-120b
Synthetic Stack Exchange from gpt-oss-120bTextUndisclosedStack Exchangegpt-oss-120b
Synthetic GPQA from gpt-oss-120b and Qwen2.5-32B-InstructTextUndisclosedStack Exchangegpt-oss-120b; Qwen2.5-32B-Instruct
Synthetic Vedantu from gpt-oss-120bTextUndisclosedVedantugpt-oss-120b
Synthetic SWE-Gym and R2E-Gym-Subset from Qwen3-Coder-480B-A35B-InstructTextUndisclosedSWE-Gym; R2E-Gym-SubsetQwen3-Coder-480B-A35B-Instruct
Synthetic SWE-Gym from Qwen3-Coder-480B-A35B-InstructTextUndisclosedSWE-GymQwen3-Coder-480B-A35B-Instruct
Synthetic SWE-Gym and R2E-Gym-Subset from DeepSeek-R1-0528TextUndisclosedSWE-Gym; R2E-Gym-SubsetDeepSeek-R1-0528
Synthetic HelpSteer, LMSYS-Chat-1M, and Nemotron-Personas-USA from gpt-oss-120b, Qwen3-235B-A22B-Instruct-2507, and Qwen3-235B-A22B-Thinking-2507TextUndisclosedHelpSteer2; HelpSteer3; LMSYS-Chat-1M; Nemotron-Personas-USAgpt-oss-120b; Qwen3-235B-A22B-Instruct-2507; Qwen3-235B-A22B-Thinking-2507
Synthetic Structured Outputs from Qwen3-30B-A3B-Instruct-2507, Qwen3-30B-A3B-Thinking-2507, Qwen3-235B-A22B-Instruct-2507, and Qwen3-235B-A22B-Thinking-2507TextUndisclosed-Qwen3-30B-A3B-Instruct-2507; Qwen3-30B-A3B-Thinking-2507; Qwen3-235B-A22B-Instruct-2507; Qwen3-235B-A22B-Thinking-2507
Synthetic Search STEM MCQ from Qwen3-235B-A22B and DeepSeek-R1-0528TextUndisclosed-Qwen3-235B-A22B; DeepSeek-R1-0528
Synthetic Search STEM OPENQ from DeepSeek-R1-0528TextUndisclosed-DeepSeek-R1-0528
Synthetic OpenSTEM from Qwen2.5-32B-Instruct and DeepSeek-R1-0528TextUndisclosed-Qwen2.5-32B-Instruct; DeepSeek-R1-0528
Synthetic MCQ from Qwen2.5-32B-Instruct and DeepSeek-R1-0528TextUndisclosed-Qwen2.5-32B-Instruct; DeepSeek-R1-0528
Synthetic MCQ10 from DeepSeek-R1-0528TextUndisclosed-DeepSeek-R1-0528
Synthetic MCQ4 from Qwen3-235B-A22B, DeepSeek-R1-0528, and Qwen3-235B-A22B-Instruct-2507TextUndisclosed-Qwen3-235B-A22B; DeepSeek-R1-0528; Qwen3-235B-A22B-Instruct-2507
Synthetic OpenMathReasoning from gpt-oss-120b and Qwen2.5-32B-InstructTextUndisclosedOpenMathReasoninggpt-oss-120b; Qwen2.5-32B-Instruct
Synthetic Offline Search MCQA HLE from DeepSeek-R1-0528TextUndisclosed-DeepSeek-R1-0528
Synthetic Offline Search MCQA GPQA from Qwen3-235B-A22B and DeepSeek-R1-0528TextUndisclosed-Qwen3-235B-A22B; DeepSeek-R1-0528
Synthetic Human Preference from QwQ-32B, Qwen3-30B-A3B, Qwen3-235B-A22B, Qwen3-235B-A22B-Instruct-2507, Mistral-Small-3.1-24B-Instruct-2503, Mistral-Small-3.2-24B-Instruct-2506, MiniMax-M1-80k, MiniMax-M1-40k, Kimi-K2-Instruct, DeepSeek-V3-0324, DeepSeek-R1-0528TextUndisclosed—QwQ-32B; Qwen3-30B-A3B; Qwen3-235B-A22B; Qwen3-235B-A22B-Instruct-2507; Mistral-Small-3.1-24B-Instruct-2503; Mistral-Small-3.2-24B-Instruct-2506; MiniMax-M1-80k; MiniMax-M1-40k; Kimi-K2-Instruct; DeepSeek-V3-0324; DeepSeek-R1-0528
Synthetic WildChat-1M and arena-human-preference-140k from DeepSeek-R1, gemma-2-2b-it, gemma-3-27b-it, gpt-oss-20b, gpt-oss-120b, Mistral-7B-Instruct-v0.3, Mixtral-8x22B-Instruct-v0.1, Nemotron-4-340B-Instruct, NVIDIA-Nemotron-Nano-9B-v2, Phi-4-mini-instruct, Phi-3-small-8k-instruct, Phi-3-medium-4k-instruct, Qwen3-235B-A22B, QwQ-32BTextUndisclosedWildChat-1M; arena-human-preference-140kDeepSeek-R1; gemma-2-2b-it; gemma-3-27b-it; gpt-oss-20b; gpt-oss-120b; Mistral-7B-Instruct-v0.3; Mixtral-8x22B-Instruct-v0.1; Nemotron-4-340B-Instruct; NVIDIA-Nemotron-Nano-9B-v2; Phi-4-mini-instruct; Phi-3-small-8k-instruct; Phi-3-medium-4k-instruct; Qwen3-235B-A22B; QwQ-32B
Synthetic Safety from DeepSeek-R1-0528, gpt-oss-120b, DeepSeek-R1-Distill-Qwen-7B, and Mixtral-8x7B-v0.1TextUndisclosedNemotron Content Safety Dataset V2; Gretel Synthetic Safety Alignment Dataset; RedTeam-2K; Malicious TasksDeepSeek-R1-0528; gpt-oss-120b; DeepSeek-R1-Distill-Qwen-7B; Qwen3-30B-A3B-Thinking-2507; Qwen3-235B-A22B-Instruct-2507; Mixtral-8x7B-v0.1
Synthetic Code from Qwen3-32BTextUndisclosedEnglish Common Crawl; English Common Crawl 1.1Qwen3-32B
Synthetic OpenCodeReasoning from DeepSeek-R1TextUndisclosedOpenCodeReasoningDeepSeek-R1
Synthetic LIMO from DeepSeek-R1-0528TextUndisclosedLIMODeepSeek-R1-0528
Synthetic SCP from DeepSeek-R1-0528TextUndisclosedSCP-116KDeepSeek-R1-0528
Synthetic Stack Exchange from DeepSeek-R1-0528TextUndisclosedStack ExchangeDeepSeek-R1-0528
Synthetic Common Crawl from Qwen3-30B-A3BTextUndisclosedCommon CrawlQwen3-30B-A3B
Synthetic Wikipedia from Qwen3-30B-A3BTextUndisclosedWikimediaQwen3-30B-A3B
Synthetic Essential-Web from Qwen3-30B-A3B and Qwen3-235B-A22B-Thinking-2507TextUndisclosedEssential-WebQwen3-30B-A3B; Qwen3-235B-A22B-Thinking-2507
Synthetic Textbook Math from Qwen3-30B-A3B, Qwen3-235B-A22B, phi-4TextUndisclosedCommon Crawl; FineMathQwen3-30B-A3B; Qwen3-235B-A22B; phi-4
Synthetic Math and Code from DeepSeek-R1 and DeepSeek-R1-0528TextUndisclosedMagicoder-Evol-Instruct-110K; opc-sft-stage2; TACO; OpenCodeReasoning; OpenMathReasoning; NuminaMath CoTDeepSeek-R1; DeepSeek-R1-0528
Synthetic Nemotron-Personas-USA from gpt-oss-120b and Qwen3-8BTextUndisclosedNemotron-Personas-USAgpt-oss-120b; Qwen3-8B

Training Dataset Token Counts

Dataset# of Tokens in Nemotron Nano 2# of Tokens in Nemotron Nano 3
English Common Crawl3,360,110,334,8183,456,523,212,210
English Synthetic CC1,949,464,641,1234,340,740,677,920
Crawl++360,389,153,262360,389,153,262
Math124,606,230,663154,217,502,165
Synthetic Math73,007,767,15573,007,767,155
Code747,409,228,7241,043,856,922,136
Synthetic Code175,067,553,293453,117,917,176
Common Crawl Code0263,072,374,097
English Wiki17,349,266,92617,349,266,926
Synthetic Wiki07,850,648,552
Books00
Papers191,586,493,365191,586,493,365
PDF-to-text141,096,578,533141,096,578,533
Code SFT60,025,726,817102,863,752,325
STEM SFT272,680,426,295359,826,214,274
General SFT6,057,478,6456,057,478,645
Tool-Calling SFT026,244,716,867
Multilingual2,172,261,909,3501,743,892,490,859
Synthetic multilingual997,710,364,950595,140,661,135
Total10,648,823,153,91913,336,833,827,602
We use a considerable amount of synthetic data. Out of 10.6 trillion tokens, 3,534,013,958,278 tokens are synthetically generated. We extracted data for fifteen languages from the following three Common Crawl snapshots: CC-MAIN-2024-51, CC-MAIN-2025-08, CC-MAIN-2025-18. The fifteen languages included were Arabic, Chinese, Danish, Dutch, French, German, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Swedish, and Thai. As we did not have reliable multilingual model-based quality classifiers available, we applied just heuristic filtering instead—similar to what we did for lower quality English data in the Nemotron-CC pipeline, but selectively removing some filters for some languages that did not work well. Deduplication was done in the same way as for Nemotron-CC. Additionally, we used data from Wikipedia and FineWeb-2 (Penedo et al., 2025) for these fifteen languages as well as four additional languages: Czech, Finnish, Hebrew, and Hindi.
LanguageTotal Tokens
Arabic118,056,362,726
Danish117,747,321,618
German146,613,691,781
Spanish469,156,575,409
French139,982,002,289
Italian298,858,370,174
Japanese682,755,693,336
Korean127,099,747,538
Dutch89,041,592,681
Polish105,356,493,147
Portuguese243,249,275,089
Russian185,314,014,057
Swedish74,954,953,299
Thai160,778,944,467
Chinese211,007,236,689
We collect a total of 922,476,782,017 tokens of code in 43 different languages.
LanguageTokens
Assembly750,628,764
C42,657,300,868
C#56,153,329,307
C++67,773,701,658
CommonLisp263,234,672
CSS38,848,760,035
Cuda400,222,993
Dart3,816,960,470
Dockerfile474,958,084
Fortran1,105,049,387
Go8,332,419,480
Haskell1,294,613,669
HTML69,082,117,487
Java131,440,465,822
JavaScript75,573,420,861
JSON15,366,881,241
Julia621,046,949
JupyterNotebook2,241,893,197
Lua4,146,420,802
Makefile12,640,010,879
Markdown64,796,743,311
Mathematica320,504,225
OmniversePython26,946,093
Pascal1,625,013,876
Perl1,575,314,434
PHP61,575,339,005
Python126,916,727,384
R19,811,381,935
reStructuredText1,779,876,391
Ruby6,446,962,615
Rust4,438,640,533
Scala3,343,959,154
Shell18,758,779,250
SQL23,205,633,085
Swift5,976,714,881
SystemVerilog233,056,185
TeX7,347,157,527
TypeScript15,657,838,582
Verilog811,884,369
VHDL648,401,444
VisualBasic.NET1,005,680,881
XML12,616,779,741
YAML10,574,010,491

Language Distribution in Post-Training

For our post-training recipe, we focused on 5 main languages in addition to English: Spanish, French, Japanese, Italian, German. Those languages were represented in the form of multilingual reasoning and translation task. The following table depicts our sample distribution for the 6 languages and 5 translation pairs.
Language / Translation PairSize
English16.2M
Italian0.252M
German0.252M
Spanish0.252M
French0.252M
Japanese0.252M
English ↔ Italian108k
English ↔ German108k
English ↔ Spanish108k
English ↔ French108k
English ↔ Japanese108k

Evaluation Dataset

PropertyValue
Data Collection MethodHybrid: Human, Synthetic
Labeling MethodHybrid: Automated, Human, Synthetic
We evaluated our model on the following benchmarks:
TaskNVIDIA-Nemotron-3-Nano-30B-A3B-BF16Qwen3-30B-A3B-Thinking-2507GPT-OSS-20B
General Knowledge
MMLU-Pro78.380.975.0
Reasoning
AIME25 (no tools)89.185.091.7
AIME25 (with tools)99.2-98.7
GPQA (no tools)73.073.471.5
GPQA (with tools)75.0-74.2
LiveCodeBench (v6 2025-08-2025-05)68.366.061.0
SciCode (subtask)33.333.034.0
HLE (no tools)10.69.810.9
HLE (with tools)15.5-17.3
MiniF2F pass@150.05.712.1
MiniF2F pass@3279.916.843.0
Agentic
Terminal Bench (hard subset)8.55.06.0
SWE-Bench (OpenHands)38.822.034.0
TauBench V2 (Airline)48.058.038.0
TauBench V2 (Retail)56.958.838.0
TauBench V2 (Telecom)42.226.349.7
TauBench V2 (Average)49.047.748.7
BFCL v453.846.4*-
Chat & Instruction Following
IFBench (prompt)71.551.065.0
Scale AI Multi Challenge38.544.833.8
Arena-Hard-V2 (Hard Prompt)72.149.6*71.2*
Arena-Hard-V2 (Creative Writing)63.266.0*25.9&
Arena-Hard-V2 (Average)67.757.848.6
Long Context
AA-LCR35.959.034.0
RULER-100@256k92.989.4-
RULER-100@512k91.384.0-
RULER-100@1M86.377.5-
Multilingual
MMLU-ProX (avg over langs)59.577.6*69.1*
WMT24++ (en->xx)86.285.683.2
Model Specifications
LicenseCustom
Last UpdatedJanuary 2026
Input TypeText
Output TypeText
ProviderNvidia
Languages6 Languages
Quick Start
Sign in to get startedView DocumentationAzure Support