Search AI Products and News

DeepSeek affiliate company publicly disclosed a patent for deploying large language models, innovatively utilizing a distributed architecture: deploying the prefill and decode stages on high-performance computing and large-memory machines respectively. This method balances the load, reduces computational idle time, significantly lowering latency and increasing throughput. The patent highlights optimizations in system scalability and fault tolerance. Its MoE language model, DeepSeek-V3, has 671 billion parameters, with 37 billion parameters activated per token, which will promote the practical application of AI technology across industries. The core breakthrough lies in combining hardware resources effectively.

San Francisco AI startup Deep Cogito launches the Cogito v2 series of open-source large models, focusing on machine intuition and self-improvement capabilities. The series includes four dense and mixture-of-experts models with 70B to 671B parameters. The flagship 671B MoE model has a reasoning efficiency that is 60% higher than DeepSeek R1, and its performance is close to Claude4Opus. Its core technology involves internalizing reasoning paths into model weights through introspective reasoning, enabling continuous evolution while in use. Tests show that the model can

The paper by Dr. Liang Wenfeng's team at DeepSeek in collaboration with Peking University has won the ACL 2025 Best Paper Award. This research proposes the Native Sparse Attention (NSA) mechanism, which achieves algorithm-hardware co-optimization through three parallel attention branches: compression, selection, and sliding. It improves the processing speed of long texts by 11 times and outperforms traditional full-attention models. NSA supports a context length of up to 1 million tokens, achieving an 11.6 times decoding speed improvement on 64k text, and a 27B parameter model demonstrates excellent performance on multiple benchmark tests.

Quest Mobile reports top 20 fastest-growing apps (June 2025). AIGC's DeepSeek leads with 163M MAU, Doubao hits 141M (+410% YoY). Hongguo (212M) and UnionPay (188M) lead short videos & payments. List reflects AIGC and online video growth potential.....

DeepSeek's Q2 downloads plunged 72.2% to 22.58M, losing market appeal. Doubao emerged as new leader while Kimi declined. Experts cite lost cost edge, cheap APIs, user churn, weak use cases, and R2 delays. Meanwhile, 'AI+Office' grew 190.2% and 'AI+Education' hit 10M users, signaling intense competition demanding constant innovation.....

The newly launched AI search function in WeChat has attracted widespread attention due to allegations of leaking personal privacy. Recently, many users reported on social platforms that this function can generate a personal resume with a name hyperlink in one click, causing concerns about privacy security among users. According to user feedback, the controversy surrounding WeChat AI Search mainly focuses on its automatic identification mechanism. When users encounter names in WeChat official account articles, the system automatically converts the name into a blue hyperlink. Clicking this link will force the AI system to generate a detailed information page containing personal resume, as well as display all

Recently, the tech brand Fansang introduced its new portable private cloud product - FX2510 Smart NAS at the 2025 São Paulo Consumer Electronics and Appliance Exhibition (Eletrolar Show Brazil 2025). This upgraded NAS not only offers powerful storage capabilities but also deeply integrates interfaces of more than ten mainstream AI models, including ChatGPT and Deepseek, fully demonstrating the wide application of smart technology in the storage field. FX

According to a report by Bloomberg, in January this year, the low-cost and high-performance AI model launched by Chinese AI company DeepSeek caused a shock in the global technology community and brought significant urgency to China's tech giant Alibaba. To quickly catch up with this technological breakthrough, engineers at Alibaba canceled their vacation during the most important traditional Chinese festival — the Spring Festival — choosing to stay overnight in the company and fully commit themselves to AI R&D. Chairman of Alibaba, Joseph Tsai, vividly told this story of racing against time at the VivaTech Science and Technology Conference held in Paris on Wednesday.

Global renowned open-source AI platform Hugging Face recently released the open-weight model contributor ranking. Chinese teams Qwen and DeepSeek successfully ranked in the top 15, showcasing China's technical strength and influence in the global open-source AI field. This ranking honors the teams that provide high-quality model weights to the open-source community, with their models widely applied in academic and industrial innovation. The Qwen team supported by Alibaba Cloud Intelligence Group has gained favor from the community due to the excellent performance of its Qwen3 series models in tasks such as instruction following and code generation.

Recently, Sam Altman, CEO of OpenAI, announced on social platform X that the open source model originally scheduled for release in early summer will now be delayed until late summer. This news came as a surprise to industry insiders who were all looking forward to the launch. However, Altman explained that this delay was due to some "unexpected and quite impressive" achievements made by the team during the R&D process. In his post, Altman mentioned that OpenAI wants to spend more time on the weights of the open source model to ensure the best possible outcome.

The field of artificial intelligence continues to develop rapidly, and DeepSeek's latest update for its R1-0528 model has drawn significant attention from the developer community. According to AIbase, the upgraded DeepSeek R1-0528 now supports local tool calls on OpenRouter, a major improvement that enables developers to seamlessly integrate local tools into their AI workflows. Local tool calls enhance workflow efficiency. DeepSeek R1-0528 is an open-source inference model that rivals Ope

A new study led by Apple researchers has poured cold water on the current large reasoning models (LRMs) that are highly anticipated. The research found that when solving complex tasks, reasoning models designed to simulate thinking processes, such as Claude3.7Thinking and Deepseek-R1, not only failed to show advantages but also exhibited serious problems like insufficient reasoning and performance collapse. This study tested four classic logical puzzles: the Tower of Hanoi, Checkers, River Crossing, and Block World. These puzzles allow precise control over the tasks.

Google recently released its latest preview version of Gemini2.5Pro, a product known as the 'most intelligent' model, which was first announced in March and then updated in May. The preview version aims to give enterprises an opportunity to test new applications or replace earlier versions. According to Google's official blog, this new version performs better in responsiveness and creativity, while surpassing other models in coding and reasoning ability. At the I/O Developer Conference in May, Google revealed that Gemini2.

Silicon Valley investor Mary Meeker's latest released AI report reveals that the global AI competition landscape is experiencing profound reshaping. Chinese strength and open source wave are shaking up the ecosystem dominated by top companies like OpenAI, entering a new stage of 'contested supremacy'. The report points out that China's large language models are rapidly catching up with the U.S., with models such as DeepSeek, Alibaba Qwen, and Baidu Wenxin Yiyan showing excellent performance in core tasks such as code generation and mathematical reasoning.