OpenAI’s biggest rival, DeepSeek, is said to be setting up a stage to compete against GPT-5 by launching a new AI model by the end of the year. The company is expected to be working on two differentiators: stronger agent capabilities and adaptive learning.
According to people familiar with the project, DeepSeek’s upcoming model will be designed to carry out multi-step tasks with minimal user input. Instead of requiring constant prompts, it’s expected to function more autonomously.
A key feature under development is the model’s ability to adapt based on its own past actions. This means the system could refine how it executes tasks over time, offering responses and workflows better tailored to context. If successful, this would be a game-changer from the static behavior of current AI systems and move toward something closer to self-improvement.
If DeepSeek is truly planning to launch a GPT-5 rival by the end of 2025, then the timing couldn’t be better. As the much-awaited GPT-5 has received cooler responses from the AI community. As a result, OpenAI CEO Sam Altman has already confirmed that GPT-6 will arrive sooner than expected, indicating that the company is eager to reset momentum.
Altman has suggested that GPT-6 will adapt more deeply to users rather than simply respond to them. He also highlighted memory as the next major milestone, making AI tools “truly personal” while stressing the importance of privacy safeguards, especially for sensitive domains like healthcare and law.
So all these things hint that if DeepSeek lives up to its promises of a self-improving, autonomous AI system, it could carve out a strong position in the market and put real pressure on OpenAI. However, with GPT-6 already in the works, the competition in advanced AI models is set to intensify going forward.
