- DeepSeek released a minor update to its R1 reasoning model (R1-0528), now available on its chatbot platforms and Hugging Face, though specific changes were not disclosed.
- Despite limited details, R1-0528 gained attention for improved coding performance, ranking as the top Chinese model on LiveCodeBench, ahead of major competitors like Alibaba’s Qwen3 and Anthropic’s Claude 3.7.
- As DeepSeek prepares for its next major release, R2, competition in China’s AI sector continues to intensify, with rivals like Alibaba and Baidu advancing in open-source rankings and multimodal AI development.
Chinese AI start-up DeepSeek has released an updated version of its R1 reasoning model. This is the first update since the model’s launch in January. The new version, called R1-0528, is available on DeepSeek’s chatbot website, mobile apps, and the open-source platform Hugging Face. The company announced the update in a WeChat group but did not share specific details.
DeepSeek R1-0528 Update and Market Context
DeepSeek did not respond to requests for comment on the R1-0528 update. The company’s last major update was in March, when it released V3-0324 with improvements in coding and writing. While the details of the R1-0528 update remain undisclosed, it has drawn strong interest from the developer community.
According to benchmark platform LiveCodeBench, the new model shows improved AI-assisted coding performance. It now ranks as the top Chinese model for coding, trailing only OpenAI’s o4-mini-high, o3-high, and o4-mini-medium. It also surpassed Alibaba’s Qwen3 and Anthropic’s Claude 3.7, both considered strong performers in AI coding.
Expectations are rising for DeepSeek’s next major reasoning model, R2. A minor update to a specialized math-solving model in April has further fueled speculation. However, DeepSeek recently lost its position as China’s top open-source model developer. Alibaba’s Qwen3 now leads on LiveBench, which ranks models based on coding, math, data analysis, and language instruction.
As competition in China’s AI sector grows, companies are expanding beyond text-based models. Baidu introduced its Ernie 4.5 Turbo and X1 Turbo models in April, highlighting their multimodal capabilities, which DeepSeek models currently lack.
Did you subscribe to our Daily newsletter?
It’s Free! Click here to Subscribe!
Source: SouthChinaMorningPost
