Tech Desk 
May 29, 2025
DeepSeek, the Chinese artificial intelligence (AI) startup which recently rose to popularity, has launched an updated version of its R1 reasoning model, known as R1-0528, heating up competition with American rivals like OpenAI.
Although no specifics, public announcement or description was provided about the new update by DeepSeek, it was released on the developer platform Hugging Face.
In terms of code generation performance, DeepSeek's updated R1 model is falling only one step short of what it takes to outdo OpenAI's o4 mini and o3 reasoning models, according to the LiveCodeBench leaderboard, a benchmark developed by researchers from UC Berkeley, MIT, and Cornell.
The surprising aspect of the toned-up DeepSeek R1 model is that it outperformed xAI's Grok 3 mini and Alibaba's Qwen 3.
Citing a Bloomberg report, Reuters noted that a DeepSeek representative informed a WeChat group about a "minor trial upgrade," which allows users to begin testing the new model.
Earlier this year, DeepSeek's development of a series of reasoning AI models, which competed or even surpassed their counterparts from leading AI firms in the US, was an answer to the assumption that US export sanctions deterred China's AI progress.
These AI models were built on significantly less budgets than those consumed by renowned US-based startups like OpenAI.
The launch of the first DeepSeek R1 model in January led tech shares outside China to plunge and reshaped perceptions about the resources needed for AI scalability.
The Chinese tech tycoon is also expected to release DeepSeek R2, the successor to R1, after an upgrade to its V3 large language model.
