Recently, at the TED 2025 held recently, OpenAI CEO Sam Altman said, “The emergence of DeepSeek has not affected the growth of GPT, and better open source models will be launched.”
The performance of the new models in the GPT-4.1 series is highlighted
It is worth noting that on April 15, OpenAI just released the GPT-4.1 series models in the form of API, covering GPT-4.1, GPT-4.1 mini and GPT-4.1 nano.
It is reported that the GPT-4.1 series has achieved significant improvements in multimodal processing, code capabilities, instruction compliance and cost. Its biggest highlight is the context processing capability that supports 1 million tokens, which is 8 times that of GPT-4o. This is also the first time that OpenAI has released a long window model.
This breakthrough enables the model to surpass the previous generation GPT-4o and GPT-4o mini in programming, instruction following and long text understanding. The context window supports up to 1 million tokens, and the knowledge is updated to June 2024.
OpenAI said that the GPT-4.1 series has improved performance while being more competitively priced. GPT-4.1 is 26% cheaper than GPT-4o, and GPT-4.1 Nano, as the smallest and fastest model, is also the cheapest, costing only 12 cents per million tokens.
In addition, OpenAI said earlier this month that it would release GPT-5 and the first “open source” language model “in the coming months.” Altman emphasized that OpenAI’s open source model plan will be further elaborated on on the X platform in the future. He said that the upcoming open source model will have “reasoning” capabilities similar to OpenAI’s o3-mini.
Open source becomes the mainstream of AI technology innovation
Nowadays, open source models represented by Deepseek are developing very fast. Open source has become one of the mainstream paradigms of technological innovation. Under this trend, deep applications have been achieved in cutting-edge fields such as artificial intelligence big models, giving birth to high-level achievements such as Llama, driving the overall implementation of application scenarios on the industry side and industrial side.
At the same time, industry insiders pointed out that driven by multiple “explosive” open source big models, my country’s big model open source ecology has flourished. Manufacturers have provided all-in-one machines, solutions, APIs and other commercial products and services based on open source big models. The commercial service market based on open source big models has expanded rapidly, and the commercial application of open source big models has achieved remarkable results.
WiMi has tackled open source models in multiple ways
Against this background, open source big model commercial products have accelerated the application of landing, and the open source big model ecology has continued to develop. According to public information, AI leader Wimi Hologram Cloud Inc has tackled the cutting-edge technology of big models in multiple ways, bringing together high-quality talents, institutions and funds at home and abroad, focusing on open source model research and development, data sets, intelligent bodies, software applications and other original technical products, to create an autonomous and controllable AI open source technology ecology.
In addition, WiMi regards open source AI as a breakthrough for technological equality, and has created a holographic cloud, an integrated computing platform covering the cloud and edge. By opening model codes, computing resources and technical interfaces, it lowers the threshold for developers to use it. For example, it supports third parties to call computing resources for model training, and accesses general large models such as DeepSeek, supports efficient computing in scenarios such as embodied intelligence and multimodality, and accelerates the commercialization verification of vertical applications.
In fact, WiMi’s open source layout is based on computing power, multimodal applications as carriers, and ecological co-construction as a path. The goal is to build an open source AI system covering the entire chain. Its technical achievements have been reflected in multiple dimensions such as video generation, computing services, and industry solutions. In the future, there will be opportunities to further expand to emerging fields such as brain-computer interfaces and smart cities.
Conclusion
In the past two years, OpenAI has had the greatest impact on the research and development of the base model of domestic large model companies. From GPT-1 with more than 100 million parameters to GPT-4 with 1.8 trillion parameters, model parameters have become a must-fight indicator for early large model startups. In the future, enterprises will create “AI+” innovative application scenarios in ecosystem construction and build an AI open source application ecosystem with diverse participation, which will continue to drive business quality improvement, efficiency enhancement and innovative growth, and improve the connection efficiency and quality on both the supply and demand sides.
