Join Forward-thinking Leaders
Elevate your expertise with tech insights, startup breakthroughs, and leadership intelligence curated for your priorities.
An Update of R1 Reasoning Model Introduced By DeepSeek
DeepSeek, a Chinese artificial intelligence startup, issued an update to its R1 reasoning model early Thursday, stepping up the battle with US rivals such as OpenAI.
Summary:
1. The updated model, R1-0528, is just below OpenAI’s o4 mini and o3 reasoning models on code creation but ahead of xAI’s Grok 3 mini and Alibaba’s Qwen 3.
2. The reasoning model is expected to be released in May, and DeepSeek has previously published AI models that were on pace with or better than industry-leading models in the United States at just under half the cost.
3. This move increases competitiveness with US-based companies such as OpenAI and Google.
According to the latest tech news, DeepSeek released R1-0528 on the development platform Hugging Face but has yet to make an official public statement. It did not publish a description of the model and comparisons.
However, the LiveCodeBench scoreboard, established by academics from UC Berkeley, MIT, and Cornell, placed DeepSeek’s improved R1 reasoning model just below OpenAI’s o4 mini and o3 reasoning models on code creation, but ahead of xAI’s Grok 3 mini and Alibaba’s Qwen 3.
Some media agencies have announced the update on Wednesday. It mentioned that a DeepSeek employee informed a WeChat group that it had finished a “minor trial upgrade” and that users may begin testing it.
When the AI tools published AI models later this year that were on pace with or better than industry-leading models in the United States at just under half of the cost, the startup countered myths that China’s AI accomplishments were being hampered by U.S. export restrictions.
What is the Reasoning Model Update?
According to the latest tech update, the official launch of R1 in January sent tech markets outside China dropping and called into question the notion that scaling AI needed massive computer capacity and investment. Since R1’s launch, Chinese tech titans such as Alibaba and Tencent have created models that promise to outperform DeepSeek’s.
Google’s Gemini has introduced lower tiers of access to it, while OpenAI has reduced pricing and developed an O3 Mini model that uses fewer processors.
The business is still widely expected to release R2, a follow-up to R1. According to insiders, R2’s release date was originally scheduled for May. The AI tool also announced an improvement to its V3 large language model in March.
Also, recently, Microsoft has partnered with Yotta Data Services, India’s leading provider of sovereign cloud infrastructure and platform services, to promote AI adoption. According to global AI news, Microsoft will deploy its AzureAI services to Shakti Cloud, its AI cloud platform, as part of this one-of-a-kind collaboration, bringing cutting-edge artificial intelligence (AI) features to developers, startups, businesses, and public sector organisations across India.
The cooperation offers an additional benefit: a) Azure AI’s advanced AI models, apps, agents, and AI development environment; and b) Yotta’s sovereign, cost-effective, globally compared AI generation platform, that allows fast AI model training and real-time prediction in critical sectors such as agriculture, healthcare, education, finance, manufacturing, retail, and media.
According to this tech news, the partnership also helps to advance the IndiaAI Mission by increasing local AI skills, fostering innovation, and improving AI infrastructure. By May 2025, the IndiaAI Mission had received over 500 submissions for developing indigenous AI models. Microsoft and Yotta will work closely with government agencies, academic institutes, IITs, and startups to foster domestic innovation and boost the development of indigenous AI models suitable with India’s Digital Public Infrastructure.
By letting Microsoft’s cutting-edge AI models and services to run on Yotta’s sovereign world-class AI infrastructure, the partnership expects to pave the way for hybrid AI that prioritises safety, trust, and low-latency AI innovation, allowing developers and enterprises to build and scale AI responsibly while maintaining data sovereignty.
