TECH

DeepSeek Unveils Major Upgrade to Match Rivals Like ChatGPT

Published: May. 30, 2025 6:01 a.m. GMT+8

00:00

00:00/00:00

Listen to this article 1x

A user interface message on the DeepSeek artificial intelligence app on a mobile phone on Feb. 28, 2025.

Chinese AI startup DeepSeek has released a major upgrade to its flagship language model, DeepSeek R1 — marking the first significant update in five months — claiming its overall capabilities now approach its major rivals.

The company quietly launched the new version, DeepSeek-R1-0528, on open-source platform Hugging Face on Wednesday morning, followed by a formal announcement later in the day detailing its capabilities and availability.

You've accessed an article available only to subscribers

Subscribe today for just $.99.

VIEW OPTIONS

Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.

Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.

Share this article

Open WeChat and scan the QR code

DIGEST HUB

Digest Hub

Back

Explore the story in 30 seconds

DeepSeek released a major upgrade, R1-0528, to its language model, achieving 87.5% accuracy on the 2025 American Invitational Mathematics Examination, up from 70%, and halving AI hallucinations in key tasks.
The model and its smaller variant, Qwen3-8B, are accessible via DeepSeek’s platforms; Qwen3-8B reportedly matches much larger models’ performance.
Tencent quickly integrated R1-0528 into its tool ima, highlighting DeepSeek's growing influence in China’s competitive AI landscape.

AI generated, for reference only

Who’s Who

DeepSeek: DeepSeek is a Chinese AI startup known for its advanced language model, DeepSeek R1. The latest upgrade, DeepSeek-R1-0528, significantly improves reasoning, reduces hallucinations, and produces more human-like writing. It rivals leading models like GPT-4o and Gemini 2.5 Pro. DeepSeek has rapidly influenced China’s AI ecosystem, with major firms integrating its models and prompting widespread competition in AI development across the country.

OpenAI: The article mentions that DeepSeek’s upgraded R1 model now approaches the capabilities of top AI models such as OpenAI’s GPT-4o. OpenAI is a leading artificial intelligence research organization known for developing advanced large language models, with GPT-4o being its latest state-of-the-art offering.

Google: According to the article, Google’s Gemini 2.5 Pro is mentioned as a top-tier language model. DeepSeek’s upgraded model, DeepSeek-R1-0528, claims to have improved reasoning abilities and depth of thought, bringing its capabilities closer to those of leading models like OpenAI’s GPT-4o and Google’s Gemini 2.5 Pro. No further details about Google are provided in the article.

Tencent: Tencent has integrated DeepSeek’s latest model into its personal knowledge management tool, "ima," just hours after the model’s open-source release. Ima now supports both Tencent’s proprietary Hunyuan model and DeepSeek, with weekly updates planned. This marks one of the fastest AI model deployments by Tencent.

Alibaba: Alibaba is mentioned as one of the major Chinese tech firms that released their own AI models to compete with DeepSeek R1 following its viral debut. This surge in AI development was part of a broader competitive response within China’s tech ecosystem, with companies like Alibaba, Baidu, Tencent, and ByteDance all entering the field to rival DeepSeek’s advancements.

Baidu: According to the article, Baidu is one of several Chinese firms—including Alibaba, Tencent, and ByteDance—that have released their own AI models to compete with DeepSeek R1 following its debut.

ByteDance: According to the article, ByteDance is one of several major Chinese companies that have released their own AI models to compete with DeepSeek R1, following its debut and rapid rise in popularity.

Huawei: The article mentions Huawei only briefly, stating that DeepSeek founder Liang Wenfeng attended a high-level roundtable with Chinese President Xi Jinping in February, alongside several top entrepreneurs, including Huawei’s Ren Zhengfei. No further details about Huawei’s involvement with DeepSeek or its AI models are provided.

AI generated, for reference only

What Happened When

December 2023：: DeepSeek launched its first reasoning model, R1.

January 2024：: DeepSeek R1 went viral, drawing global attention to China’s AI industry.

February 2024：: DeepSeek founder Liang Wenfeng was invited to a high-level roundtable with Chinese President Xi Jinping, along with other top entrepreneurs.

December 2024：: DeepSeek V3 base model was released.

May 28, 2025：: DeepSeek quietly launched the new version, DeepSeek-R1-0528, on open-source platform Hugging Face.

May 28, 2025：: DeepSeek made a formal announcement about DeepSeek-R1-0528, detailing its capabilities and availability.

May 28, 2025：: Tencent announced the integration of DeepSeek R1-0528 into its 'ima' tool, hours after the open-source release.

As of May 28, 2025：: Users could access the upgraded model via DeepSeek’s website, app, or WeChat mini program, and 'Deep Thinking' feature was available.