DeepSeek to launch GPT-5 rival later this year: Here's what we know so far

The measure that came into force on Saturday aims to block new downloads of the app, the agency said, though DeepSeek's web service remains accessible in the country.

Chinese artificial intelligence startup DeepSeek is preparing to release a new AI model by the end of 2025. The model is being positioned as a direct competitor to OpenAI’s GPT-5, with a particular emphasis on enhanced agent features and adaptive learning.

Features

According to people familiar with the project, the upcoming DeepSeek model is designed to carry out multi-step tasks on behalf of users with minimal direction. Unlike conventional systems that rely on frequent prompts, this model is being developed to operate more autonomously, enabling smoother interactions and reducing user effort.

A distinguishing factor is the model’s planned ability to learn and adapt from its own past actions. This adaptive mechanism is expected to improve the AI’s efficiency over time, allowing it to refine how it completes assignments and respond more effectively to different contexts. Such a feature represents a notable shift from static AI systems, potentially setting a new standard in real-world usability.

Competition

DeepSeek’s initiative reflects China’s broader strategy to strengthen its technological capabilities and compete with leading US firms in AI. While OpenAI continues to dominate the global landscape with its GPT line of models, DeepSeek is positioning itself as an emerging challenger by focusing on agent-driven automation and self-improving systems.

Industry analysts note that the development of autonomous and adaptive AI is becoming increasingly important. Businesses and consumers alike are seeking systems that can handle complex workflows without constant human oversight, a demand that companies like OpenAI, Anthropic, and now DeepSeek are racing to meet.