Yandex has introduced its latest generation of language models, YandexGPT 4, featuring two versions: Lite and Pro. These new generative neural networks surpass earlier models, particularly in handling longer and more complex queries.
YandexGPT 4 can now process approximately 60 pages of text, making it capable of solving more diverse and challenging tasks.
Test results show that YandexGPT 4 Pro delivers higher-quality answers than previous models, with significant improvements in 70% of cases. For tasks require
ng open-ended responses, the model approaches the performance level of GPT-4. It also offers enhanced reasoning abilities, which enable it to handle questions that involve analysis.
Compared to the third-generation models, YandexGPT 4 processes four times more text and delivers answers twice as fast.
This capability allows it to address longer and more complicated queries efficiently. Additionally, YandexGPT 4 can generate commands for third-party applications. For example, it can create a command for an airline ticket service to launch a seat search function.
The YandexGPT 4 models are now available via API on Yandex Cloud, although some features are still in testing. Users can try out the new models in a demo chat on the Yandex Cloud platform. Yandex plans to integrate these models into its services in the future.
YandexGPT 4 follows the launch of YandexGPT 3, which debuted earlier this year.
Source: Yandex Press Release
