Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2604.24954 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918493410033664 |
|---|---|
| author | NVIDIA : Deshmukh, Amala Sanjay Chumachenko, Kateryna Rintamaki, Tuomas Le, Matthieu Poon, Tyler Taheri, Danial Mohseni Karmanov, Ilia Liu, Guilin Seppanen, Jarno Goel, Arushi Ranzinger, Mike Heinrich, Greg Chen, Guo Voegtle, Lukas Fischer, Philipp Roman, Timo Sapra, Karan McCarthy, Collin Zhang, Shaokun Liu, Fuxiao Ye, Hanrong Dong, Yi Liu, Mingjie Peng, Yifan Zelasko, Piotr Chen, Zhehuai Koluguri, Nithin Rao Tadevosyan, Nune Grigoryan, Lilit Asl, Ehsan Hosseini Biswas, Pritam Tavabi, Leili Su, Yuanhang Yu, Zhiding Jin, Peter Milesi, Alexandre Haber, Netanel Xu, Yao Amiraslani, Sarah Mulepati, Nabin Tramel, Eric Jung, Jaehun Lu, Ximing Cui, Brandon Xu, Jin Li, Zhiqi Wang, Shihao Kuang, Yuanguo Zhang, Shaokun Yang, Huck Li, Boyi Yin, Hongxu Han, Song Kartal, Bilal Molchanov, Pavlo Renduchintala, Adi Wang, Charles Mosallanezhad, David Singhal, Soumye Vega, Luis Cheung, Katherine Ghosh, Sreyan Zhang, Yian Bukharin, Alexander Srinivasan, Venkat Greco, Johnny Manoel, Andre Van Segbroeck, Maarten Panguliri, Suseella Watve, Rohit Kakwani, Divyanshu Pachori, Shubham Glick, Jeffrey Sri-Tharan, Radha Zaman, Aileen Nguyen, Khanh Chen, Shi Fang, Jiaheng Miao, Qing Zhou, Wenfei Wang, Yu Bhat, Zaid Pervaiz Praveen, Varun Jain, Arihant Arunachalam, Ramanathan Kornuta, Tomasz Sharabiani, Ashton Shen, Amy Huang, Wei Wu, Yi-Fu Ghias, Ali Roshan Li, Huiying Yu, Brian Tajbakhsh, Nima Cui, Chen Gao, Wenwen Ding, Li Kong, Terry Kilaru, Manoj Bhiwandiwalla, Anahita Wawrzos, Marek Korzekwa, Daniel Ribalta, Pablo Chlebus, Grzegorz Nushi, Besmira Dobrowolska, Ewa Mikulski, Maciej Jakub Dhawan, Kunal Huang, Steve Balam, Jagadeesh Wang, Yongqiang Karpov, Nikolay Mendelev, Valentin Zelenfroynd, George Mkrtchyan, Meline Miao, Qing Almog, Omri Pawar, Bhavesh Shivbhakta, Rameshwar Sabnis, Sudeep Sharabiani, Ashrton Habibi, Negar Venkataramani, Geethapriya Peng, Pamela Rodney, Prerit Panev, Serge Mazzarese, Richard Liu, Nicky Fukuyama, Michael Skliar, Andrii Waleffe, Roger Riach, Duncan Zou, Yunheng Hu, Jian Zhang, Hao Xu, Binfeng Yang, Yuhao Ahmed, Zuhair Milesi, Alexandre del Mundo, Carlo Voegele, Chad Cheng, Zhiyu Assaf, Nave Skliar, Andrii Afrimi, Daniel Bagrov, Natan Zilberstein, Ran Masad, Ofri Khvedchenia, Eugene Bagrov, Natan Tymchenko, Borys Asida, Tomer Afrimi, Daniel Mannan, Parth Cui, Victor Evans, Michael Luna, Katherine Lou, Jie Xu, Pinky Huang, Guyue Habibi, Negar Boone, Michael Thalasta, Pradeep Adesoba, Adeola Yared, Dina Parisien, Christopher Derczynski, Leon Ghosh, Shaona Feely, Wes Schaffer, Micah Sri-Tharan, Radha Glick, Jeffrey Simkin, Barnaby Zelenfroynd, George Grzegorzek, Tomasz Garg, Rishabh Jhunjhunwala, Aastha Kolchenko, Sergei Memarian, Farzan Kumar, Haran Kumar, Shiv Hulseman, Isabel Shah, Anjali Briski, Kari Subramanian, Padmavathy Conway, Joey Karpas, Udi Scowcroft, Jane Polak Surla, Annie Ammireddy, Shilpa Evans, Ellie Oliver, Jesse Balough, Tom Chen, Chia-Chih Bhaskar, Sandip Rico, Alejandra Sadeghi, Bardiya Mard, Seph Cheung, Katherine Price, Meredith Sleiman, Laya Kaji, Saori Helmholz, Wesley Quan, Wendy Lightstone, Michael Cohen, Jonathan Zhang, Jian Kuchaiev, Oleksii Ginsburg, Boris Kautz, Jan Long, Eileen Shoeybi, Mohammad Patwary, Mostofa Olabiyi, Oluwatobi Tao, Andrew Catanzaro, Bryan Karpas, Udi |
| author_facet | NVIDIA : Deshmukh, Amala Sanjay Chumachenko, Kateryna Rintamaki, Tuomas Le, Matthieu Poon, Tyler Taheri, Danial Mohseni Karmanov, Ilia Liu, Guilin Seppanen, Jarno Goel, Arushi Ranzinger, Mike Heinrich, Greg Chen, Guo Voegtle, Lukas Fischer, Philipp Roman, Timo Sapra, Karan McCarthy, Collin Zhang, Shaokun Liu, Fuxiao Ye, Hanrong Dong, Yi Liu, Mingjie Peng, Yifan Zelasko, Piotr Chen, Zhehuai Koluguri, Nithin Rao Tadevosyan, Nune Grigoryan, Lilit Asl, Ehsan Hosseini Biswas, Pritam Tavabi, Leili Su, Yuanhang Yu, Zhiding Jin, Peter Milesi, Alexandre Haber, Netanel Xu, Yao Amiraslani, Sarah Mulepati, Nabin Tramel, Eric Jung, Jaehun Lu, Ximing Cui, Brandon Xu, Jin Li, Zhiqi Wang, Shihao Kuang, Yuanguo Zhang, Shaokun Yang, Huck Li, Boyi Yin, Hongxu Han, Song Kartal, Bilal Molchanov, Pavlo Renduchintala, Adi Wang, Charles Mosallanezhad, David Singhal, Soumye Vega, Luis Cheung, Katherine Ghosh, Sreyan Zhang, Yian Bukharin, Alexander Srinivasan, Venkat Greco, Johnny Manoel, Andre Van Segbroeck, Maarten Panguliri, Suseella Watve, Rohit Kakwani, Divyanshu Pachori, Shubham Glick, Jeffrey Sri-Tharan, Radha Zaman, Aileen Nguyen, Khanh Chen, Shi Fang, Jiaheng Miao, Qing Zhou, Wenfei Wang, Yu Bhat, Zaid Pervaiz Praveen, Varun Jain, Arihant Arunachalam, Ramanathan Kornuta, Tomasz Sharabiani, Ashton Shen, Amy Huang, Wei Wu, Yi-Fu Ghias, Ali Roshan Li, Huiying Yu, Brian Tajbakhsh, Nima Cui, Chen Gao, Wenwen Ding, Li Kong, Terry Kilaru, Manoj Bhiwandiwalla, Anahita Wawrzos, Marek Korzekwa, Daniel Ribalta, Pablo Chlebus, Grzegorz Nushi, Besmira Dobrowolska, Ewa Mikulski, Maciej Jakub Dhawan, Kunal Huang, Steve Balam, Jagadeesh Wang, Yongqiang Karpov, Nikolay Mendelev, Valentin Zelenfroynd, George Mkrtchyan, Meline Miao, Qing Almog, Omri Pawar, Bhavesh Shivbhakta, Rameshwar Sabnis, Sudeep Sharabiani, Ashrton Habibi, Negar Venkataramani, Geethapriya Peng, Pamela Rodney, Prerit Panev, Serge Mazzarese, Richard Liu, Nicky Fukuyama, Michael Skliar, Andrii Waleffe, Roger Riach, Duncan Zou, Yunheng Hu, Jian Zhang, Hao Xu, Binfeng Yang, Yuhao Ahmed, Zuhair Milesi, Alexandre del Mundo, Carlo Voegele, Chad Cheng, Zhiyu Assaf, Nave Skliar, Andrii Afrimi, Daniel Bagrov, Natan Zilberstein, Ran Masad, Ofri Khvedchenia, Eugene Bagrov, Natan Tymchenko, Borys Asida, Tomer Afrimi, Daniel Mannan, Parth Cui, Victor Evans, Michael Luna, Katherine Lou, Jie Xu, Pinky Huang, Guyue Habibi, Negar Boone, Michael Thalasta, Pradeep Adesoba, Adeola Yared, Dina Parisien, Christopher Derczynski, Leon Ghosh, Shaona Feely, Wes Schaffer, Micah Sri-Tharan, Radha Glick, Jeffrey Simkin, Barnaby Zelenfroynd, George Grzegorzek, Tomasz Garg, Rishabh Jhunjhunwala, Aastha Kolchenko, Sergei Memarian, Farzan Kumar, Haran Kumar, Shiv Hulseman, Isabel Shah, Anjali Briski, Kari Subramanian, Padmavathy Conway, Joey Karpas, Udi Scowcroft, Jane Polak Surla, Annie Ammireddy, Shilpa Evans, Ellie Oliver, Jesse Balough, Tom Chen, Chia-Chih Bhaskar, Sandip Rico, Alejandra Sadeghi, Bardiya Mard, Seph Cheung, Katherine Price, Meredith Sleiman, Laya Kaji, Saori Helmholz, Wesley Quan, Wendy Lightstone, Michael Cohen, Jonathan Zhang, Jian Kuchaiev, Oleksii Ginsburg, Boris Kautz, Jan Long, Eileen Shoeybi, Mohammad Patwary, Mostofa Olabiyi, Oluwatobi Tao, Andrew Catanzaro, Bryan Karpas, Udi |
| contents | We introduce Nemotron 3 Nano Omni, the latest model in the Nemotron multimodal series and the first to natively support audio inputs alongside text, images, and video. Nemotron 3 Nano Omni delivers consistent accuracy improvements over its predecessor, Nemotron Nano V2 VL, across all modalities, enabled by advances in architecture, training data and recipes. In particular, Nemotron 3 delivers leading results in real-world document understanding, long audio-video comprehension, and agentic computer use. Built on the highly efficient Nemotron 3 Nano 30B-A3B backbone, Nemotron 3 Nano Omni further incorporates innovative multimodal token-reduction techniques to deliver substantially lower inference latency and higher throughput than other models of similar size. We are releasing model checkpoints in BF16, FP8, and FP4 formats, along with portions of the training data and codebase to facilitate further research and development. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_24954 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence NVIDIA : Deshmukh, Amala Sanjay Chumachenko, Kateryna Rintamaki, Tuomas Le, Matthieu Poon, Tyler Taheri, Danial Mohseni Karmanov, Ilia Liu, Guilin Seppanen, Jarno Goel, Arushi Ranzinger, Mike Heinrich, Greg Chen, Guo Voegtle, Lukas Fischer, Philipp Roman, Timo Sapra, Karan McCarthy, Collin Zhang, Shaokun Liu, Fuxiao Ye, Hanrong Dong, Yi Liu, Mingjie Peng, Yifan Zelasko, Piotr Chen, Zhehuai Koluguri, Nithin Rao Tadevosyan, Nune Grigoryan, Lilit Asl, Ehsan Hosseini Biswas, Pritam Tavabi, Leili Su, Yuanhang Yu, Zhiding Jin, Peter Milesi, Alexandre Haber, Netanel Xu, Yao Amiraslani, Sarah Mulepati, Nabin Tramel, Eric Jung, Jaehun Lu, Ximing Cui, Brandon Xu, Jin Li, Zhiqi Wang, Shihao Kuang, Yuanguo Zhang, Shaokun Yang, Huck Li, Boyi Yin, Hongxu Han, Song Kartal, Bilal Molchanov, Pavlo Renduchintala, Adi Wang, Charles Mosallanezhad, David Singhal, Soumye Vega, Luis Cheung, Katherine Ghosh, Sreyan Zhang, Yian Bukharin, Alexander Srinivasan, Venkat Greco, Johnny Manoel, Andre Van Segbroeck, Maarten Panguliri, Suseella Watve, Rohit Kakwani, Divyanshu Pachori, Shubham Glick, Jeffrey Sri-Tharan, Radha Zaman, Aileen Nguyen, Khanh Chen, Shi Fang, Jiaheng Miao, Qing Zhou, Wenfei Wang, Yu Bhat, Zaid Pervaiz Praveen, Varun Jain, Arihant Arunachalam, Ramanathan Kornuta, Tomasz Sharabiani, Ashton Shen, Amy Huang, Wei Wu, Yi-Fu Ghias, Ali Roshan Li, Huiying Yu, Brian Tajbakhsh, Nima Cui, Chen Gao, Wenwen Ding, Li Kong, Terry Kilaru, Manoj Bhiwandiwalla, Anahita Wawrzos, Marek Korzekwa, Daniel Ribalta, Pablo Chlebus, Grzegorz Nushi, Besmira Dobrowolska, Ewa Mikulski, Maciej Jakub Dhawan, Kunal Huang, Steve Balam, Jagadeesh Wang, Yongqiang Karpov, Nikolay Mendelev, Valentin Zelenfroynd, George Mkrtchyan, Meline Miao, Qing Almog, Omri Pawar, Bhavesh Shivbhakta, Rameshwar Sabnis, Sudeep Sharabiani, Ashrton Habibi, Negar Venkataramani, Geethapriya Peng, Pamela Rodney, Prerit Panev, Serge Mazzarese, Richard Liu, Nicky Fukuyama, Michael Skliar, Andrii Waleffe, Roger Riach, Duncan Zou, Yunheng Hu, Jian Zhang, Hao Xu, Binfeng Yang, Yuhao Ahmed, Zuhair Milesi, Alexandre del Mundo, Carlo Voegele, Chad Cheng, Zhiyu Assaf, Nave Skliar, Andrii Afrimi, Daniel Bagrov, Natan Zilberstein, Ran Masad, Ofri Khvedchenia, Eugene Bagrov, Natan Tymchenko, Borys Asida, Tomer Afrimi, Daniel Mannan, Parth Cui, Victor Evans, Michael Luna, Katherine Lou, Jie Xu, Pinky Huang, Guyue Habibi, Negar Boone, Michael Thalasta, Pradeep Adesoba, Adeola Yared, Dina Parisien, Christopher Derczynski, Leon Ghosh, Shaona Feely, Wes Schaffer, Micah Sri-Tharan, Radha Glick, Jeffrey Simkin, Barnaby Zelenfroynd, George Grzegorzek, Tomasz Garg, Rishabh Jhunjhunwala, Aastha Kolchenko, Sergei Memarian, Farzan Kumar, Haran Kumar, Shiv Hulseman, Isabel Shah, Anjali Briski, Kari Subramanian, Padmavathy Conway, Joey Karpas, Udi Scowcroft, Jane Polak Surla, Annie Ammireddy, Shilpa Evans, Ellie Oliver, Jesse Balough, Tom Chen, Chia-Chih Bhaskar, Sandip Rico, Alejandra Sadeghi, Bardiya Mard, Seph Cheung, Katherine Price, Meredith Sleiman, Laya Kaji, Saori Helmholz, Wesley Quan, Wendy Lightstone, Michael Cohen, Jonathan Zhang, Jian Kuchaiev, Oleksii Ginsburg, Boris Kautz, Jan Long, Eileen Shoeybi, Mohammad Patwary, Mostofa Olabiyi, Oluwatobi Tao, Andrew Catanzaro, Bryan Karpas, Udi Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition We introduce Nemotron 3 Nano Omni, the latest model in the Nemotron multimodal series and the first to natively support audio inputs alongside text, images, and video. Nemotron 3 Nano Omni delivers consistent accuracy improvements over its predecessor, Nemotron Nano V2 VL, across all modalities, enabled by advances in architecture, training data and recipes. In particular, Nemotron 3 delivers leading results in real-world document understanding, long audio-video comprehension, and agentic computer use. Built on the highly efficient Nemotron 3 Nano 30B-A3B backbone, Nemotron 3 Nano Omni further incorporates innovative multimodal token-reduction techniques to deliver substantially lower inference latency and higher throughput than other models of similar size. We are releasing model checkpoints in BF16, FP8, and FP4 formats, along with portions of the training data and codebase to facilitate further research and development. |
| title | Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence |
| topic | Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.24954 |