_version_ 1866913889212432384
author GigaChat team
Valentin, Mamedov
Kosarev, Evgenii
Leleytner, Gregory
Shchuckin, Ilya
Berezovskiy, Valeriy
Smirnov, Daniil
Kozlov, Dmitry
Averkiev, Sergei
Ivan, Lukyanenko
Proshunin, Aleksandr
Israfilova, Ainur
Baskov, Ivan
Chervyakov, Artem
Shakirov, Emil
Kolesov, Mikhail
Khomich, Daria
Latortseva, Darya
Porkhun, Sergei
Fedorov, Yury
Kutuzov, Oleg
Kudriavtseva, Polina
Soldatova, Sofiia
Egor, Kolodin
Pyatkin, Stanislav
Menshykh, Dzmitry
Sergei, Grafov
Damirov, Eldar
Vladimir, Karlov
Gaitukiev, Ruslan
Shatenov, Arkadiy
Fenogenova, Alena
Savushkin, Nikita
Minkin, Fedor
author_facet GigaChat team
Valentin, Mamedov
Kosarev, Evgenii
Leleytner, Gregory
Shchuckin, Ilya
Berezovskiy, Valeriy
Smirnov, Daniil
Kozlov, Dmitry
Averkiev, Sergei
Ivan, Lukyanenko
Proshunin, Aleksandr
Israfilova, Ainur
Baskov, Ivan
Chervyakov, Artem
Shakirov, Emil
Kolesov, Mikhail
Khomich, Daria
Latortseva, Darya
Porkhun, Sergei
Fedorov, Yury
Kutuzov, Oleg
Kudriavtseva, Polina
Soldatova, Sofiia
Egor, Kolodin
Pyatkin, Stanislav
Menshykh, Dzmitry
Sergei, Grafov
Damirov, Eldar
Vladimir, Karlov
Gaitukiev, Ruslan
Shatenov, Arkadiy
Fenogenova, Alena
Savushkin, Nikita
Minkin, Fedor
contents Generative large language models (LLMs) have become crucial for modern NLP research and applications across various languages. However, the development of foundational models specifically tailored to the Russian language has been limited, primarily due to the significant computational resources required. This paper introduces the GigaChat family of Russian LLMs, available in various sizes, including base models and instruction-tuned versions. We provide a detailed report on the model architecture, pre-training process, and experiments to guide design choices. In addition, we evaluate their performance on Russian and English benchmarks and compare GigaChat with multilingual analogs. The paper presents a system demonstration of the top-performing models accessible via an API, a Telegram bot, and a Web interface. Furthermore, we have released three open GigaChat models in open-source (https://huggingface.co/ai-sage), aiming to expand NLP research opportunities and support the development of industrial solutions for the Russian language.
format Preprint
id arxiv_https___arxiv_org_abs_2506_09440
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GigaChat Family: Efficient Russian Language Modeling Through Mixture of Experts Architecture
GigaChat team
Valentin, Mamedov
Kosarev, Evgenii
Leleytner, Gregory
Shchuckin, Ilya
Berezovskiy, Valeriy
Smirnov, Daniil
Kozlov, Dmitry
Averkiev, Sergei
Ivan, Lukyanenko
Proshunin, Aleksandr
Israfilova, Ainur
Baskov, Ivan
Chervyakov, Artem
Shakirov, Emil
Kolesov, Mikhail
Khomich, Daria
Latortseva, Darya
Porkhun, Sergei
Fedorov, Yury
Kutuzov, Oleg
Kudriavtseva, Polina
Soldatova, Sofiia
Egor, Kolodin
Pyatkin, Stanislav
Menshykh, Dzmitry
Sergei, Grafov
Damirov, Eldar
Vladimir, Karlov
Gaitukiev, Ruslan
Shatenov, Arkadiy
Fenogenova, Alena
Savushkin, Nikita
Minkin, Fedor
Computation and Language
Artificial Intelligence
Generative large language models (LLMs) have become crucial for modern NLP research and applications across various languages. However, the development of foundational models specifically tailored to the Russian language has been limited, primarily due to the significant computational resources required. This paper introduces the GigaChat family of Russian LLMs, available in various sizes, including base models and instruction-tuned versions. We provide a detailed report on the model architecture, pre-training process, and experiments to guide design choices. In addition, we evaluate their performance on Russian and English benchmarks and compare GigaChat with multilingual analogs. The paper presents a system demonstration of the top-performing models accessible via an API, a Telegram bot, and a Web interface. Furthermore, we have released three open GigaChat models in open-source (https://huggingface.co/ai-sage), aiming to expand NLP research opportunities and support the development of industrial solutions for the Russian language.
title GigaChat Family: Efficient Russian Language Modeling Through Mixture of Experts Architecture
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2506.09440