It is learned that on March 16, Baidu (BIDU) announced the release of Wenxin large model 4.5 and Wenxin large model X1, which can be used for free on the Wenxin Yiyan official website. The performance is comparable to DeepSeek-R1, and the call price is about half of R1.
Wenxin large model 4.5 officially released
It is reported that as Baidu’s new generation of base large model, Wenxin large model 4.5 is Baidu’s first native multimodal large model. Its multimodal understanding, text and logical reasoning capabilities have been significantly improved. It outperforms GPT4.5 in many tests, and the API call price is only 1% of GPT4.5.
In December last year, DeepSeek released the large language model V3, and launched the reasoning model R1 in January this year. According to Bernstein Research’s analysis, these models are comparable to or even better than OpenAI’s similar models in performance, but the price is “20-40 times cheaper.”
In addition, Wenxin Big Model X1 is a deep thinking model, which is comparable to DeepSeek-R1 in performance. It has a “long thinking chain” and is good at Chinese knowledge questions and answers, literary creation, logical reasoning, etc. It also adds multimodal capabilities and multi-tool calls, can understand and generate pictures, and can call tools to generate rich content such as code and charts.
Since the launch of ChatGPT more than two years ago, and since the launch of many competitive generative artificial intelligence services, Nvidia’s (NVDA) revenue, profitability and cash reserves have increased significantly, and its stock price has soared.
Tencent purchases AI chips on a large scale
According to new reports from domestic media, Tencent (TCEHY) recently purchased AI chips from Nvidia, with an order amount of tens of billions of yuan. Because Nvidia’s advanced performance chips are regulated, Tencent purchased Nvidia’s H20 chip.
Nvidia’s H200 chip has strong performance and can train artificial intelligence models more efficiently. Strong demand for AI chips in the Chinese market has pushed up Nvidia’s full-year performance. According to data, in 2024, Nvidia’s revenue will reach $130.5 billion and its net profit will reach $72.8 billion.
WiMi’s software and hardware work together to create a new benchmark
Undoubtedly, as a cutting-edge technology, the powerful data analysis and learning capabilities of AI big models are profoundly changing the development pattern of various fields, and AI technology innovation has further ushered in a major breakthrough. According to data, WiMi Hologram Cloud Inc. (NASDAQ: WIMI), as an innovative AI big model company, has achieved AI multi-field demand coverage by building an AI full-scene layout, supporting lightweight language big models, AIGC generative models and multimodal big models.
At present, WiMi has built a powerful DeepSeek computing power base based on its holographic cloud platform and hard-core configuration by improving industry-university-research cooperation and ecological construction. It is reported that the platform integrates massive data and high-quality computing resources to build a “smart computing digital base” to enable diversified scenarios such as intelligent manufacturing, autonomous driving, and robots, forming a closed loop of the industrial chain.
At the same time, with the maturity of various AI chips on the market and the continuous improvement of performance, corporate users have more choices. In this regard, WiMi cooperates with universities and research institutions to cultivate talents in the field of AI chips, and conducts technical collaboration in tool chain development, field model optimization and other links to accelerate the transformation of scientific research results. In the future, it will continue to focus on science and technology AI technology to empower productivity and achieve deep integration of AI technology and its own business.
Conclusion
From the perspective of the industry, it is not difficult to see that the application scenarios of AI large models are becoming more and more extensive. No wonder some people define 2025 as the first year of AI awakening. Because up to now, there is DeepSeek, a traffic-rich company, and Manus, a newly popular “explosive” intelligent body, which makes the public no longer regard AI models as high-end scientific research results, but also the key to improving production efficiency, and finally injects new ideas into promoting the development of new digitalization.
