Sber has created an improved version of the neural network model GigaChat. According to the developers, it has become even more convenient, functional, and efficient. Artificial intelligence has become more competent in various fields, can maintain a conversation in several stages, and respond faster.
The GigaChat MAX neural network has learned to better highlight the semantic parts of the text, and the structure of the response has become clearer.
Specialists have updated and supplemented the datasets for training, which has led to an improvement in the quality of model responses in areas such as economics, law, and medicine. The most noticeable progress has been made in modeling scientific questions in the fields of biology, chemistry, and physics.
In addition, the quality of solving mathematical problems using the open datasets GSM8K and MATH has improved by 25%.
Sber has expanded access to the GigaChat chatbot in Telegram for all users. Previously, an invitation was required to use the neural network, but now all users can take advantage of it.
Earlier, www1.ru reported that Domclick has launched a new service — real estate selection using artificial intelligence GigaChat. The neural network will offer the best solutions in the shortest possible time. The user can formulate new questions or change the conditions, without limiting themselves in the number of requests.
Read materials on the topic:
Sber and Sibur will develop a neural network for import substitution of production equipment
From concept to video in a few minutes: Sber presented the beta version of Kandinsky Video 1.1
The first real Russian competitor to ChatGPT? Sber presented GigaChat