Chinese Startup, DeepSeek, Launches a Powerful Update to its AI Model, Now on Par with Proprietary Giants OpenAI and Google
Earlier this year, the global AI community was abuzz with the initial release of DeepSeek - an open-source reasoning AI model by the Chinese startup DeepSeek. As a spinout from the Hong Kong quantitative analysis firm High-Flyer Capital Management, DeepSeek was about to set some serious benchmarks in the world of artificial intelligence.
Following up on their initial success, DeepSeek has released an updated version of its open-source AI model, DeepSeek-R1-0528. The update represents a significant improvement over the initial model, essentially bringing it to the brink of parity with closed-source proprietary models like OpenAI’s o3 and Google Gemini 2.5 Pro.
Designed for complex reasoning tasks in diverse domains such as mathematics, sciences, business, and programming, DeepSeek-R1-0528 also offers enhanced features for developers and researchers. The model remains available under a permissive open MIT license, supporting both commercial usage and developer customization.
In addition to the impressive improvements and accessible pricing model, DeepSeek is also making serious strides in the user experience department with several new features introduced in the DeepSeek-R1-0528 update. Among these are support for JSON output and the ability to call functions, both handy tools for developers looking to integrate the model’s capabilities into their applications and workflows.
DeepSeek’s commitment to better serving its user base is evident with these enhancements, including the reduction of the model’s hallucination rate, which brings more consistency to output. They’ve also introduced system prompts to streamline the deployment process.
This significant update represents a major leap forward for the open-source AI community, showing that it’s not just the tech giants who can make waves in this space. If the trend continues, the future of AI just might be a more open one, ushering in a new era of accessibility, usability, and affordability.
- •QwenLong-L1 solves long-context reasoning challenge that stumps current LLMs venturebeat.com31-05-2025
- •DeepSeek R1-0528 arrives in powerful open source challenge to OpenAI o3 and Google Gemini 2.5 Pro venturebeat.com30-05-2025