Sber Launches GigaChat 2.0: A New Era of Neural Networks in Russian

The updated neural network model surpasses foreign counterparts in Russian language processing

The Sberbank team has announced the launch of the new generation of GigaChat 2.0 neural networks. This was reported by the bank's press service.

The GigaChat 2 Max model demonstrated outstanding results, surpassing many foreign counterparts in the field of Russian language processing. According to MERA benchmark data, GigaChat 2 Max holds a leading position among AI models, and in international tests it outperforms systems such as GPT-4o and LLaMA 70B.

GigaChat 2 Pro maintains high quality, similar to the previous Max version, and is suitable for tasks requiring creativity and accuracy. This neural network model is less demanding on resources.

The basic GigaChat 2 Lite model is designed for solving simple tasks and is comparable in quality to Pro, which allows it to handle more complex requests faster.

Users can continue to use first-generation models and test GigaChat 2.0 before transitioning to the new generation.

Read more on the topic:

GigaChat Max neural network successfully certified in the fields of finance, banking and insurance

GigaChat from Sber has mastered the ability to simplify complex texts for a wide audience

Take the Unified State Exam? Sberbank's GigaChat neural network will "enroll" in university!

Источники
Press-sluzhba Sberbanka

Now on home