Chinese artificial intelligence startup DeepSeek released an upgrade to its R1 reasoning model on Thursday, intensifying competition with US players such as OpenAI in the field of code generation.
The new version, named R1-0528, was quietly launched on the developer platform Hugging Face without an official announcement or technical documentation.
Despite the low-profile release, the model quickly gained attention after it appeared on the LiveCodeBench leaderboard — a benchmark developed by researchers at UC Berkeley, MIT and Cornell.
The updated R1 ranked just behind OpenAI’s o4 mini and o3 models but ahead of xAI’s Grok 3 mini and Alibaba’s Qwen 3 on code generation performance.
DeepSeek representative described the release as a “minor trial upgrade” in a private WeChat group, adding that users could begin testing the model immediately.
DeepSeek first drew global attention in January when it launched its R1 model, which performed at a level comparable to leading US models while requiring significantly less computing power and cost.
The launch sent shares of AI-related firms outside China tumbling and challenged the assumption that massive investment and infrastructure are essential for competitive AI scaling.
In response, major players such as OpenAI and Google’s Gemini have since adjusted their pricing and model offerings. OpenAI released the o3 mini, while Gemini introduced discounted access tiers.
Chinese tech giants, including Alibaba (9988.HK) and Tencent (0700.HK), have also rolled out new models, with some claiming to outperform DeepSeek’s R1.
Industry observers continue to watch for the launch of R2, DeepSeek’s anticipated successor to R1. It is reported that the company had initially targeted a May release for R2.
In addition to its reasoning models, DeepSeek released an update to its V3 large language model in March.
COMMENTS
Comments are moderated and generally will be posted if they are on-topic and not abusive.
For more information, please see our Comments FAQ