While many large language model companies are still struggling with "how to monetize," UNISOUND (09678) has quietly achieved a full-chain breakthrough from "technical leadership" to "product deployment" and finally to "commercial realization." If the AGI era requires an "all-round warrior," then UNISOUND is delivering a rare, high-scoring report card with its first-class technical foundation, first-class productization capability, and first-class commercial execution. The latest announcement reveals that in 2025, UNISOUND's large model-related business revenue reached 6.0 billion to 6.2 billion yuan (Renminbi), a staggering year-on-year increase of 1057% to 1095%, and is expected to account for 48% to 53% of the company's total revenue. This signifies that what was once a technological testing ground has now become a genuine "cash cow." This is not merely a reshuffling of the financial structure; it marks the Chinese AI industry's transition from a "parameter race" into the deeper waters of "value realization."
Behind this explosive growth lie three key drivers. From our perspective, UNISOUND's surge is no accident but the result of a powerful combination of technological leadership, productization capability, and deep integration into critical application scenarios. Firstly, its core technology continues to lead the pack. The company has built a comprehensive "Shan Hai" (Mountain Sea) large model matrix, encompassing general-purpose large language models, multimodal models, and vertical industry-specific models. Among these, "Shan Hai · Zhi Yi 5.0" swept the top spots in medical agent, large language model, and multimodal categories in the authoritative MedBench 4.0 evaluation. The newly launched "Shan Hai · Zhi Yin 2.0" achieves the industry's highest level in automatic speech recognition (ASR), supporting high-precision recognition for over 30 Chinese dialects and 14 international languages. Its text-to-speech (TTS) capability can synthesize 12 dialects and 10 foreign languages with remarkable human-likeness and has pioneered natural, fluid full-duplex interaction, comprehensively refreshing the human-computer dialogue experience.
Secondly, its productization capability has improved significantly. Leveraging the unified "Shan Hai · Atlas" intelligent computing base, the company packages underlying model capabilities into standardized Agent solutions. These cover high-frequency scenarios like intelligent customer service, document review, and smart marketing, substantially reducing delivery costs and implementation cycles. Most crucially, the focus is on high-value, "critical scenarios." Instead of chasing the泛entertainment or general chatbot tracks, UNISOUND has deeply cultivated fields with extremely high demands for accuracy, compliance, and stability, such as healthcare, medical insurance, and transportation. Currently, "Shan Hai · Zhi Yi" is deployed in nearly 400 hospitals, with over 700 more in testing phases. At the Shunyi branch of Beijing Friendship Hospital, the citation rate for medical record generation is close to 90%. Concurrently, renewal rates and expansion purchase intentions continue to climb, with customers validating its commercial value with real financial commitments.
Anchored as the "Leader in Professional Large Models and Agents," standing at this new starting point where large model revenue constitutes over half of its income, UNISOUND's strategic direction for 2026 is clear and resolute: to become the leader in China's professional large model and Agent domain. The company will persistently deepen its "One Base, Two Wings" technical strategy—using the self-developed "Shan Hai · Atlas" as the unified intelligent foundation to systematically advance four core capabilities: solidifying the general large model base, developing expert-level Agents, building industry-specific data flywheels, and optimizing edge-side inference and chip deployment. On this foundation, it will accelerate the deployment of vertical Agents across thousands of industries and millions of households.
Specifically, the company will leverage its "Shan Hai · Zhi Yi" system, already covering 400 hospitals, to explore expansion into a wider range of medical institutions and deepen applications in scenarios like medical insurance cost control. Simultaneously, utilizing the highly human-like interaction capabilities of "Shan Hai · Zhi Yin," it will continue crafting AI interaction portals that "understand accurately, speak naturally, and respond quickly," successfully deploying them in millions of endpoints like smart cockpits and intelligent transportation systems. More importantly, UNISOUND will systematically replicate the successful model validated in healthcare—"Industry Large Model + Agent Platform + Scenario-Specific Solutions"—to other critical, high-complexity domains such as medical insurance, government affairs, and transportation. Through the modular combination and rapid adaptation of Agents, it aims to achieve a leap from "single-point projects" to "platform-level products," truly unleashing the scaled commercial potential of large models.
In summary, while the industry still debates whether large models are a bubble, UNISOUND has demonstrated with its revenue constituting nearly half the total that the value of AI lies not in parameter scale, but in the ability to solve real-world problems. The transformation of UNISOUND might well be a microcosm of the entire Chinese AI industry. As the淘汰赛 intensifies, the market no longer pays for concepts but only for tangible value. In 2026, as more professional Agents take root in critical scenarios, UNISOUND is not only poised to reach an inflection point towards profitability but also to establish a replicable, sustainable, and profitable commercialization paradigm for the Chinese AI industry. For investors, the case of UNISOUND offers a crucial insight: in the latter half of AI investing, focus should be on companies with controllable technology, clear application scenarios, and visible revenue streams. With its unique strategic resolve and execution efficiency, UNISOUND is charting an evolutionary path for the market from "AI concept stock" to "AI value stock."
Comments