Moscow, 13 March (IANS). According to the MERA benchmark data for Russia, the Giga Chat 2 Max model of the leading global financial institution Sber is ranked first in the AI models. This updated product line compared to the international benchmark performs better than GPT-4O, Dipcic-V3, LLAMA 70B and Quen 2.5 in several matrix.
The entire Giga Chat 2.0 has been upgraded by significantly changing the product line, which has enabled business customers to solve current tasks and implement large -scale projects rapidly and better.
Giga Chat 2 Max has become even more powerful and performs better than many similar foreign models in solving the tasks prescribed in Russian language.
The Giga Chat 2 Pro shows the same quality as the previous Max version and can perform functions where creativity and accuracy is important. Also, the company said that the model requires less resources.
The original model for light tasks, Giga Chat 2 Lite, is now equal to the quality of the previous Pro version and can do complex functions rapidly and more cost effectively.
Users will still have access to the first generation model and can try Giga Chat 2.0 before upgrading to the new product line.
“You do not need to know a lot about programming to find Giga Chat and its best version of your business, Senior Vice President and Technology Development Head and Technology Development Head and Technology Development Head.” He said that Giga Chat 2.0 has not only increased metric and technical features, but is an important step in the development of the Laj language model (LLM) of the Russian-language.
“We have created a model at the level of the world’s best solutions, and in Russian-language functions, the model performs better than most of them. The strong Russian neural networks are strategically important for any company working in Russia,” Bellevatsev said.
Bellevatsev stated that 15,000 external customers already use giga chats, and this powerful upgrade in our product line will allow even more customers to solve a wide range of tasks more efficiently.
Belevatsev said, “By improving processes with the help of Artificial Intelligence, companies will have a unique opportunity to stay forward in competition, increase profit and improve customer loyalty.”
Depending on the Giga Chat 2.0, companies will be able to create more productive autonomous assistants (AI agents) that can cause and solve complex, multi-compatible problems on their own. This has been possible because models have increased their knowledge about mathematics, science and humanities and have learned to do better codes and write.
According to the company, to develop agents in python and GS, you can use popular langchen SDK, with which Giga Chat is fully compatible. Compatibility packages are available in Gigachat Repository. The next generation models maintain the context of the conversation for a very long time, answer complex, long questions and analyze more texts.
Previously, a single querry could have about 48A4 page text (14 point fonts), but now the maximum query volume has increased by about 200 pages. This makes it easier to make a chatbot with Giga Chat 2.0.
The new models are twice as accurate in following the instructions of the users and 25 percent better in answering questions, they follow specified formats and conditions and create answers in a certain style, which helps to resolve tasks such as auxiliary legal documents and analyzing customer requests.
According to the independent MERA benchmark for Russian language, the Giga Chat 2 Max ranks first in the AI model, and according to the results of MMLU benchmarking in Russian and English, the new product line is equal to or even ahead of the world’s top protesters.
The most impressive results were achieved by the leading model of the series. Compared to Deepcic-V3, Queen 2.5 (Queen-2.5-75B), GPT-4O and LLAMA 70B, Giga Chat 2 Max gives better answers to factual questions in Russian and follows the given format. According to the Humanistic Benchmark for performance in code generation works, this model performs better than its foreign counterparts and is more efficient in accurate science.
-IANS
SCH/Ekde
