Beijing's status as a leading hub for artificial intelligence innovation is being reinforced by a series of new breakthroughs from its domestic companies.
On July 16, the Beijing-based company Moonshot AI unveiled KimiK3, the world's first open-source model with a scale of 3 trillion parameters, designed for advanced intelligent scenarios like long-context programming, knowledge work, and reasoning.
According to the introduction, KimiK3 is a 2.8 trillion parameter model. Its development focus was not merely on scaling up parameter count, but on synergistic optimization of model architecture and training methods to better translate scale advantages into enhanced capabilities.
A representative from Moonshot AI stated that parameters are akin to neural connections in the human brain. Nearly 3 trillion parameters mean the model can store more knowledge and patterns, enabling it to know more, think deeper, and answer more accurately.
In evaluations released by Moonshot AI, KimiK3 demonstrated globally leading comprehensive capabilities. The model can not only complete specific code generation and modification tasks but also engage in open-ended creation and complex knowledge production.
In digital content generation scenarios, it can combine 3D reasoning, programming, and visual capabilities to transform concepts, images, or videos into interactive experiences.
In scenarios like financial research and industry analysis, it can organize and analyze corporate documents, industry data, and long-term industrial materials to produce readable, verifiable, and traceable in-depth reports.
The Moonshot AI representative revealed that the company is currently working closely with inference partners and open-source maintainers to align technical details, ensuring the model can be reliably deployed across the ecosystem.
The complete model weights are scheduled for release by July 27, 2026. More details on architecture, training, and evaluation will be disclosed alongside the KimiK3 technical report.
Scaling Up to Global Leadership
"When a model's parameters reach the trillion level, it essentially enters the global first tier," remarked Shao Bin, President of the Zhongguancun Artificial Intelligence Research Institute, who views the release of Moonshot's massive model as highly significant.
For a considerable period, the "large" in large models has primarily signified scale expansion, following the scaling law—where continuous increases in computing power, model size, and data volume lead to corresponding leaps in intelligence levels.
With ongoing investment in computing power and intensified competition in specific sectors, future models may see their parameter counts expand even further into the trillions.
Beijing's ability to support the development of one of the world's largest models by parameter count is solid proof that the city is operating at a world-class level, from model training to output generation.
Open-Source Initiatives and Market Firsts
Beijing has consistently been at the forefront of AI development: it has produced a series of internationally leading models like GLM-5.2 and KimiK3; it has registered and launched 259 large models, the highest number nationwide; and it is the birthplace of the "world's first publicly listed large model company."
This year, Zhipu AI successfully listed on the Hong Kong stock exchange, becoming that pioneering entity.
The company's President, Wang Shaolan, explained that "enabling machines to think like humans" has been Zhipu's guiding principle since its founding.
In 2025, Zhipu successively released GLM-4.5, GLM-4.6, GLM-4.7, and GLM-5, with all four generations of its foundational models ranking among the top globally for open-source models, gradually attaining reasoning capabilities approaching human levels.
In June 2026, GLM-5.2 was released, ranking first in coding capability among global open-source models. Concurrently, its user base worldwide has continued to grow, marking a rapid development phase for model applications.
On July 17, at the 2026 World Artificial Intelligence Conference (WAIC) Meteorology Session, the China Meteorological Administration launched the "Wind Harmony" large language model for AI meteorological services and initiated its global open-source plan.
"Wind Harmony" is the world's first hundred-billion-parameter open-source meteorological large language model, co-developed by the China Meteorological Administration Public Meteorological Service Center and Zhipu AI, among others. It provides intelligent support for weather analysis, risk assessment, and meteorological services.
The launch of the "Wind Harmony" global open-source plan represents not just the opening of model code, but a complete technical delivery solution, allowing global users to embed "Wind Harmony" into various terminals like embodied intelligence, apps, and mini-programs, thereby building an open, shared AI meteorology ecosystem to make weather AI services accessible to everyone.
"This case in the public meteorological service sector demonstrates that Zhipu's technology has proven effective in real-world scenarios, which in itself is a positive validation of its application value," said Wang Shaolan.
Currently, the GLM series of large models is already applied in over 20 industries including public governance, industrial manufacturing, energy and power, finance, and the internet, accelerating the march towards the AGI era.
Wang Shaolan believes that global research teams and researchers should maintain an open and collaborative attitude towards technology, rather than keeping it closed off. While continued effort is needed for technological breakthroughs, international cooperation is equally essential in areas like safety governance, model interpretability, and inherently embedding safety into models.
Replicable AI Application Experience
In the field of artificial intelligence, Beijing has established a tripartite development pattern of continuous technological innovation, highly concentrated talent resources, and deepening industrial implementation.
Shao Bin explained that in terms of technological innovation, breakthroughs come from both academic forces like universities and research institutes, and from corporate entities like Zhipu AI and Moonshot AI, with all parties continuously driving technological progress.
Simultaneously, talent is a key factor in AI competition. Beijing possesses the nation's densest concentration of research institutes and national strategic scientific research power, giving it a significant advantage in talent. Furthermore, the AI industry chain in Beijing is very complete, providing excellent ecosystem support for AI to empower countless industries.
It is worth noting that the application form of the large model industry is evolving from past conversational modes towards action-oriented modes that genuinely begin to intervene in production processes.
It is understood that in key sectors like healthcare, education, government affairs, and industry, Beijing has developed a batch of replicable and promotable AI application experiences.
Among these, Beijing has taken the lead in constructing an "AI + New Materials" innovation paradigm, with a cluster of integrated innovation achievements in "AI+New Materials" accelerating their emergence, providing reference and models for related industry development.
President Shao Bin of the Zhongguancun AI Research Institute stated that a core reason Beijing's in-depth promotion of specialized work like "AI+New Materials" can succeed is that it truly deeply binds technology demanders with technology suppliers.
By injecting real-world scenario data, models evolve from possessing general capabilities to being able to complete specific professional tasks. The technology does not remain confined to the laboratory; instead, developers are directly stationed in user scenarios to understand on-site what problems exist and what kind of technology is needed to solve them.
This way, models gain both general capabilities and professional expertise, transforming a "capable" model into a "productive" tool.
A responsible person from the Beijing Municipal Development and Reform Commission stated that in the next step, Beijing will continue to improve its policy support system, optimize the innovation ecosystem, accelerate the cultivation of new forms of the intelligent economy, and speed up the construction of a global AI innovation hub with technological source power and industrial competitiveness.
Efforts will focus on application scenario development, deeply promoting the construction of application bases and benchmark scenarios, accelerating the formation of application demonstrations in key fields and industries, significantly improving the performance and adoption rate of intelligent agents and terminals, and advancing the coordinated development of the integrated chip, model, cloud, and application ecosystem.
Comments