Chinese AI startup DeepSeek has introduced a significant upgrade to its R1 reasoning model, aiming to outdo major artificial intelligence (AI) players such as OpenAI and Google.
The updated variant, R1-0528, enhances reasoning depth and inference capabilities, resolving complex queries within the blink of an eye and narrowing the limitations with OpenAI’s o3 models and Google’s Gemini 2.5 Pro.
In response, renowned Chinese tech giants Alibaba and Tencent have introduced models that claim to outperform DeepSeek’s.
Since DeepSeek R1’s release in January, it gained significant attention, impacting tech shares outside China and challenging assumptions regarding the resources needed for AI development.
DeepSeek's latest update to R1 minimises misleading outputs
The Chinese AI platform said that the update minimises misleading outputs by nearly half in tasks, including rewriting and summarising.
The model is able to perform a variety of tasks such as code generation, role-playing, and more, while offering strong results in programming, maths, and more.
In addition, the company applied the reasoning process from R1-0528 to improve Alibaba’s Qwen 3 8B Base model, enhancing its performance by over 10%. The company further highlighted the significance of this advancement for academic research of smaller AI models.
This significant update follows efforts by competitors such as Google and OpenAI to offer lower-cost, more accessible AI solutions.
Notably, the company is likely to launch a successor, R2, in the near future.
