T-Bank has introduced T-Pro 2.0, the first Russian language model with a hybrid reasoning mode. The development is available for free use in commercial projects. For example, it can be used to create AI agents. Similar solutions were previously only available from Qwen and DeepSeek.
T-Pro 2.0 uses neural networks for processing and generating content. It can analyze text and predict words and phrases. The model can break tasks down into smaller parts, build logical chains, and engage in complex analytics.
T-Pro 2.0 can be used to automate office work, process client requests, and other processes that require abstract thinking.
Compared to the previous version, the new model has significantly improved its analytical capabilities and reduced the number of errors in responses. The development requires half the computing power than Chinese counterparts, thanks to more efficient tokenization of the Russian language.
T-Bank spent about 120 million rubles on T-Pro 2.0. The model will be available in open access for adaptation and fine-tuning. The company plans to use it to improve the efficiency of its AI assistants.
Read more on the topic:
Neural networks began to analyze payments based on the habits of bank customers
