iFlytek released its semi-annual report for 2026, posting revenue of RMB 11.623 billion for the first half, up 6.52% year-on-year. The company reported a net loss of RMB 204 million, compared with a loss of RMB 239 million in the same period last year, while net profit excluding non-recurring items came in at a loss of RMB 637 million, widening from a loss of RMB 364 million a year earlier.
In the filing, iFlytek said it is proactively driving strategic business realignment in line with the principle of strengthening the consumer segment, deepening the enterprise segment, and prioritizing the government segment, while scaling back traditional project-based operations. Government-sector revenue declined 2.65% year-on-year in the first half of 2026. Sales of the iFlytek AI learning tablet fell short of expectations during the period, affected by chip and memory price hikes and intermittent supply shortages. However, the product further cemented its edge in the premium market, with its online share in the RMB 6,000-plus price bracket rising to 49% (based on data from JD.com, Tmall, and Douyin, with an even higher offline share), and its overall market share continuing to trend upward.
The RMB 270 million year-on-year decline in recurring net profit was mainly attributed to R&D spending exceeding RMB 3 billion in the first half, up more than RMB 600 million from a year earlier and accounting for over 25% of total revenue.
In the telecom operator space, iFlytek has built a differentiated portfolio of 5G New Calling products, including AI shorthand and AI companion services, helping carriers shift from "product + AI" to "AI + product" models. During the reporting period, cumulative retained paying users surpassed 10 million, supporting operators' AI+ product strategies.
At an earnings call held in the afternoon, iFlytek President Wu Xiaoru revealed that the company will officially launch a new flagship general-purpose large language model (LLM) built entirely on domestic computing power at this year's Global 1024 Developer Festival. A preliminary version will be rolled out by the end of this month.
Wu said the upcoming flagship model is expected to rank among the top tier in China in terms of coding capability and token cost-performance, while maintaining a domestic lead across all key metrics among general-purpose LLMs trained on homegrown computing infrastructure.