iFLYTEK's subsidiary, Ciyuan Xinghuo, has made the 4B and 1.7B edge-side models of its X2.5 iteration publicly available, with a focus on 1M context processing, coding capabilities, agent functionalities, and edge deployment. The company is also set to unveil the 293B version on September 7, embracing a dual-pronged approach by deploying compact models on end-user devices while preserving technological superiority through large-scale models. iFLYTEK has recently encountered a series of hurdles. Chairman Liu Qingfeng acknowledged that constraints in domestic computing resources have rendered the company somewhat marginalized in the ongoing large model revolution, which centers around coding and intelligent agents. Concurrently, Han Yuchen, Deputy General Manager of the Brand Marketing Center, was relieved of his duties, potentially linked to a prior internal whistleblowing episode.

In an effort to bolster investor confidence, the company initiated its first share repurchase program, acquiring 750,000 shares at a cost of roughly 30.02 million yuan. The overall repurchase scheme is projected to range between 100 million and 200 million yuan, earmarked for equity incentives or employee stock ownership plans. Nevertheless, the capital market's response has been lukewarm, with the stock price plummeting nearly 40% from its 52-week peak, hovering close to its nadir.

For the first half of the year, iFLYTEK posted revenue of 11.623 billion yuan, marking a 6.52% year-on-year increase. However, it incurred a net loss attributable to shareholders amounting to 204 million yuan and a non-GAAP net loss of 637 million yuan, representing a 74.88% year-on-year surge. Research and development expenditure climbed to 3.007 billion yuan, up 25.73% from the previous year, while the net cash outflow from operating activities stood at 945 million yuan.

The company is currently undergoing business realignment, phasing out traditional project-based ventures characterized by slim profit margins and protracted payment cycles. Emerging sectors such as open platforms, smart healthcare, and smart automotive have witnessed growth, whereas traditional segments like educational products, smart hardware, and enterprise AI solutions have experienced a downturn. The learning machine business, amidst an industry-wide contraction, has managed to retain users through premium offerings but has yet to chart a new growth path.

To expedite the advancement of its model business, the company intends to spin off Xinghuo as an independent entity, implementing a flat organizational structure and seeking external funding. Nonetheless, Xinghuo will continue to consolidate its financial statements, with the listed company shouldering the costs. The company must institute appropriate evaluation mechanisms to prevent swift capital depletion.