At the recent 2026 half-year performance briefing, Zhipu disclosed the progress of its next-generation large model development, clearly stating that the new product will firmly adopt the large foundation model approach and set the core goal as "supporting full-scale business calls on the day of release," aiming to fundamentally address the industry's common pain point of "large model scale but difficult practical application."
Regarding the specific technical parameters that have been widely concerned by the outside world, Zhipu did not disclose them at this briefing and stated that it will later focus on the related work of mid-training and post-training around this powerful underlying base. From the research ideas it has disclosed, the core is to fully unleash the original potential of the base model, ensuring that when the model is launched to the market, it not only has top technical indicators but can also seamlessly connect and support actual business scenarios. To this end, Zhipu has already fully initiated preliminary research on the inference side for the next-generation model, aiming to make the product run efficiently on the first day of launch and meet the real needs of large-scale commercialization.
At the same time as disclosing the R&D plan for the next generation, Zhipu also officially released its financial results for the first half of 2026. Data shows that during the period from January to June this year, the company achieved a total operating revenue of 954 million yuan, a significant increase of 399.7% compared to the same period last year; it achieved a gross profit of 252 million yuan, an increase of 163.7% year-on-year, with an overall gross profit margin of 26.4%. Meanwhile, the net profit attributable to the parent company's shareholders was -2.071 billion yuan, with the loss narrowing by 12.1% year-on-year.
Looking at Zhipu's recent strategic moves, from continuously iterating the open-source ecosystem, launching models focused on programming capabilities, to clearly outlining the R&D blueprint of the next-generation large base model in the semi-annual report, this AI company is accelerating continuously in the dual track of technological innovation and commercialization. As the large model industry gradually moves away from mere parameter competition, Zhipu's in-depth layout of "full-scale business calls" may provide a new reference sample for the engineering implementation of the entire industry.
