Moscow— Sber has officially launched GigaChat 2.0, making its powerful neural network assistant accessible to all users. The company said on Monday that the model’s capabilities have significantly improved, thanks to a new training approach that enhances its ability to analyze complex data, recognize images and audio, and respond with greater depth and accuracy.
The new version brings together all GigaChat features into a single product available across any interface, eliminating the need to switch between services. The system now includes two options: GigaChat 2 Pro and GigaChat 2 Max. While the Pro version is geared toward efficient handling of everyday tasks such as answering questions or drafting text, the Max version is designed for advanced and professional problem-solving.
According to Sber, GigaChat 2.0 can now access real-time data from the internet, analyze queries more thoroughly, and provide concise answers supported by links to sources. It finds relevant information, filters out noise, and presents users with the most accurate data. Users can now upload large documents—up to 200 A4 pages—and ask targeted questions, such as clarifying parts of a lease agreement under Russian law.
One of the most significant advancements lies in audio processing. GigaChat 2.0 can now interpret audio files directly without converting them to text, making it more effective at identifying key points and responding to user queries. It supports files up to 60 minutes long and 30 MB in size. The model can also handle voice messages, understand multiple languages, grasp complex terminology, recognize spoken accents, and filter out background noise and music.
Users can now submit links to web content, and GigaChat will extract the key information. It can summarize articles, compare coverage of a single topic across multiple sources, and even analyze images embedded in websites. The assistant can now answer questions based on video links by interpreting audio tracks, whether it’s a lecture or a video essay, and supports content in English and other languages.
Another major feature upgrade is music and song generation. Users can now generate songs of up to three minutes, with improved prompt relevance, sound quality, and even expanded support for Chinese-language lyrics. By inputting lyrics or describing a musical theme, users can create custom tracks in any genre within about a minute.
GigaChat 2.0 has also improved its ability to interpret images. It can help users choose clothing styles, solve textbook equations, or explain medical results simply by analyzing a photo or screenshot. For example, a user can upload a utility bill and ask the assistant to break down the charges.
For the first time in Russia, Sber has integrated GigaChat with smart speakers, giving them advanced conversational capabilities. The AI can now hold longer and more contextually aware conversations, switching between roles and tones based on the user’s needs. It can explain complex topics in simple terms, set alarms, play music, or issue reminders—all in a single interaction. Users can customize the assistant’s voice, communication style, and form of address, choosing from 18 different combinations.
GigaChat 2.0 is now available on the Russian digital platform MAX by VK, which combines a messenger, mini-apps, chatbot builder, online registration, and payment systems. According to the company, MAX users can use Sber’s AI to generate content, transcribe audio, summarize videos, and get accurate answers to a wide range of queries. To try it, users can search for @gigachat on the platform and follow the instructions. (Source: IANS)
