Saved in:
Bibliographic Details
Main Authors: Liu, Alexander H., Khandelwal, Kartik, Subramanian, Sandeep, Jouault, Victor, Rastogi, Abhinav, Sadé, Adrien, Jeffares, Alan, Jiang, Albert, Cahill, Alexandre, Gavaudan, Alexandre, Sablayrolles, Alexandre, Héliou, Amélie, You, Amos, Ehrenberg, Andy, Lo, Andy, Eliseev, Anton, Calvi, Antonia, Sooriyarachchi, Avinash, Bout, Baptiste, Rozière, Baptiste, De Monicault, Baudouin, Lanfranchi, Clémence, Barreau, Corentin, Courtot, Cyprien, Grattarola, Daniele, Dabert, Darius, Casas, Diego de las, Chane-Sane, Elliot, Ahmed, Faruk, Berrada, Gabrielle, Ecrepont, Gaëtan, Guinet, Gauthier, Novikov, Georgii, Kunsch, Guillaume, Lample, Guillaume, Martin, Guillaume, Gupta, Gunshi, Ludziejewski, Jan, Rute, Jason, Studnia, Joachim, Amar, Jonas, Delas, Joséphine, Roberts, Josselin Somerville, Yadav, Karmesh, Chandu, Khyathi, Jain, Kush, Aitchison, Laurence, Fainsin, Laurent, Blier, Léonard, Zhao, Lingxiao, Martin, Louis, Saulnier, Lucile, Gao, Luyu, Buyl, Maarten, Jennings, Margaret, Pellat, Marie, Prins, Mark, Poirée, Mathieu, Guillaumin, Mathilde, Dinot, Matthieu, Futeral, Matthieu, Darrin, Maxime, Augustin, Maximilian, Chiquier, Mia, Schimpf, Michel, Grinsztajn, Nathan, Gupta, Neha, Raghuraman, Nikhil, Bousquet, Olivier, Duchenne, Olivier, Wang, Patricia, von Platen, Patrick, Jacob, Paul, Wambergue, Paul, Kurylowicz, Paula, Muddireddy, Pavankumar Reddy, Chagniot, Philomène, Stock, Pierre, Agrawal, Pravesh, Torroba, Quentin, Sauvestre, Romain, Soletskyi, Roman, Menneer, Rupert, Vaze, Sagar, Barry, Samuel, Gandhi, Sanchit, Waghjale, Siddhant, Gandhi, Siddharth, Ghosh, Soham, Mishra, Srijan, Aithal, Sumukh, Antoniak, Szymon, Scao, Teven Le, Cachet, Théo, Sorg, Theo Simon, Lavril, Thibaut, Saada, Thiziri Nait, Chabal, Thomas, Foubert, Thomas, Robert, Thomas, Wang, Thomas, Lawson, Tim, Bewley, Tom, Edwards, Tom, Jamil, Umar, Tomasini, Umberto, Nemychnikova, Valeriia, Phung, Van, Maladière, Vincent, Richard, Virgile, Bouaziz, Wassim, Li, Wen-Ding, Marshall, William, Li, Xinghui, Yang, Xinyu, Ouahidi, Yassine El, Wang, Yihan, Tang, Yunhao, Ramzi, Zaccharie
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.08584
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • We introduce the Ministral 3 series, a family of parameter-efficient dense language models designed for compute and memory constrained applications, available in three model sizes: 3B, 8B, and 14B parameters. For each model size, we release three variants: a pretrained base model for general-purpose use, an instruction finetuned, and a reasoning model for complex problem-solving. In addition, we present our recipe to derive the Ministral 3 models through Cascade Distillation, an iterative pruning and continued training with distillation technique. Each model comes with image understanding capabilities, all under the Apache 2.0 license.