Determining if a medical AI company has truly crossed the commercialisation divide requires more than just examining model parameters, product counts, or press conference buzz. The real test comes from answering three practical questions: Has AI capability become the primary revenue driver? Can a single technology base continuously produce new specialist applications? And can the model achieve high-level regulatory approval to genuinely integrate into clinical workflows? The 2026 interim results from DIAGENS-B (02526), released on August 7th, provide clear answers to these questions.
During the reporting period, the company achieved total operating revenue of RMB 108.7 million, a 21.0% year-on-year increase. Gross profit reached RMB 80.504 million, up 14.0%, with a consolidated gross margin remaining high at 74.1%. Notably, revenue from model services hit RMB 94.541 million, surging 101.1% year-on-year and accounting for 86.9% of total revenue. If 2025 was a verification period for DIAGENS-B's business model transformation, then having model service revenue approach 90% of total income signals that the company has crossed a critical threshold. Its medical imaging large model is no longer just a technical foundation or an add-on feature but has become the core business driving revenue growth.
Revenue Base Shift: Model Services Contribute More Than Entire Net Growth
The substance of DIAGENS-B's revenue growth this period is primarily reflected in the composition of its increase. In the same period last year, technology licensing revenue was RMB 46.96 million, making up 52.3% of customer contract revenue. In the first half of 2026, following an adjustment, model service revenue rose to RMB 94.541 million, increasing its share of total revenue to 86.9%. Model services added approximately RMB 47.58 million in new revenue within six months, while the company's overall net revenue increased by about RMB 18.89 million. The former was 2.5 times the latter. This indicates that model services not only contributed all of the company's revenue growth but also effectively absorbed short-term fluctuations from medical imaging software and device projects delayed by budget approvals, tenders, and acceptance procedures.
DIAGENS-B's revenue base has shifted from device and software sales to model capability output. Renaming "technology licensing" to "model services" does not change the substance of contracts or revenue recognition methods. What truly matters is the expansion of business content into three complementary delivery models: model and technology licensing, the iMedMaaS® cloud service, and the SCTI localised integrated storage, computing, training, and inference appliance. These three models cater to different requirements from medical institutions regarding data security, private deployment, model training, and computing resources. Consequently, DIAGENS-B no longer sells just a pre-existing algorithm; it offers a complete production service that transforms hospital imaging data and doctor expertise into specialised AI capabilities. Although the old and new categories are not entirely comparable on the same basis, model service revenue in the first half of 2026 has already surpassed the full-year technology licensing revenue of RMB 84.34 million in 2025, clearly demonstrating the rapid expansion of this business.
The Real Core Asset is Not 158 Models, But the System That Produces Them
Viewing DIAGENS-B merely as a company with 158 specialised models would underestimate the capability demonstrated in this interim report. Traditional medical imaging AI often follows a "one disease, one model" development approach, requiring new data collection, annotation, algorithm training, and clinical validation for each new condition. This model has long development cycles, requires significant specialist input, and offers limited reusability across different projects. DIAGENS-B is tackling the fundamental challenge of scaling medical AI production. The company utilises the iMedImage® medical imaging base model for reusable image understanding and reasoning capabilities. It employs iMedStudio™ for data processing, professional annotation, manual correction, multi-person review, and quality control. It relies on iMedMaaS® for specialised model training, customisation, evaluation, release, and deployment. Finally, DoctorBench® is used to assess model capability, safety, and applicability boundaries.
The iMedLoop™, released in July 2026, further connects these capabilities, integrating data ingestion, intelligent annotation, quality control, model training, unified evaluation, release and deployment, and application feedback. This creates an end-to-end platform from medical imaging data to the application of model results. As of the announcement date, the iMedLoop™ system involved over 3,000 professionals and had accumulated approximately 28.95 million annotated samples. DoctorBench® covers three tracks (language models, multimodal models, and clinical task agents), 25 medical scenarios and tasks, and approximately 12,920 test items. Within the disclosed test sets and evaluation scope, iMedImage® ranked first overall, with its performance in eight scenario categories all placing within the top three. These data points converge on a fact more significant than "model count": DIAGENS-B has organised previously fragmented data, tools, expertise, and engineering processes into a medical imaging AI production line.
From Project Collaboration to Regulated Products, the Clinical Loop Begins to Form
As of the end of June 2026, DIAGENS-B had initiated 158 model projects, collaborating with 99 hospitals, including 65 top-tier tertiary hospitals, covering 43 human organs or application sites and 61 disease directions. These project collaborations are not equivalent to 158 commercial products or 99 paying customers. Their deeper value lies in the continuous accumulation of task definitions, annotation standards, evaluation methods, deployment processes, and clinical feedback for various specialities. Each completed project adds not just a model but a set of professional experiences reusable for subsequent tasks. For medical AI to generate genuine clinical and commercial value, it must clear the critical regulatory hurdle.
On May 19, 2026, DIAGENS-B's AI AutoVision® Chromosome Karyotype Image-Assisted Diagnostic Software obtained a Class III medical device registration certificate from the National Medical Products Administration (NMPA). This product assists in the cutting, counting, recognition, arrangement, and abnormal tip prompts for G-banded chromosome karyotype images from peripheral blood and amniotic fluid samples. The importance of this Class III certificate lies not only in adding an approved product but also in validating DIAGENS-B's ability to transform its medical imaging base model into a regulated medical device, establishing a complete path from "base model—specialist development—clinical validation—registration application—commercial delivery". Individual algorithms can be replicated, but high-quality medical data, clinical collaboration experience, regulatory registration capabilities, and hospital deployment systems require long-term accumulation. DIAGENS-B has integrated these high-barrier elements into a single production platform.
The Platform Flywheel Turns, Growth No Longer Relies on Single Products
The future growth potential of DIAGENS-B should not be estimated using a static method like "158 models multiplied by revenue per model." There is no one-to-one correspondence between model projects, diagnostic tasks, and revenue contracts; simple multiplication would obscure the true value of the platform business model. A more appropriate observation method is to see if two types of reuse can occur simultaneously. Horizontal reuse involves tasks. iMedImage® provides common underlying capabilities, while iMedStudio™, iMedMaaS®, and DoctorBench® offer standardised toolchains. New specialist tasks can reuse existing model capabilities, data governance methods, evaluation systems, and deployment experience, thereby expanding the range of organs, diseases, and imaging modalities the platform can serve.
Vertical reuse involves institutions. Medical institutions can start with model training or technology licensing, then gradually extend to local deployment, system integration, model iteration, and new tasks. The three models—cloud service, technology licensing, and local appliance—provide multi-layered entry points for institutions of different sizes and with varying data security requirements. Crucially, practical application feeds back to enhance underlying capabilities. Compliant data, test results, deployment experience, and doctor feedback generated from specialist models in clinical use can be used for model improvement and subsequent task development. This forms a closed loop: "base model—specialist model—services and products—practical application feedback—model iteration." The richer the applications, the more mature the model and engineering system become. The stronger the foundation, the higher the efficiency for developing and delivering new tasks. This cyclical accumulation is the core platform effect distinguishing DIAGENS-B from single-point medical AI product companies.
Policy Resonance and Investment, the Industrialisation Window is Opening
The external environment for medical imaging AI is also evolving. The "Implementation Opinions on Promoting and Regulating the Application of 'AI + Healthcare'" issued by five national ministries clearly states a goal: by 2030, promote the widespread application of AI technologies like intelligent medical imaging diagnosis in secondary-level hospitals and above. This establishes a clear demand-side timeline for medical imaging AI. The data supply side is also accelerating. In June 2026, the National Healthcare Security Administration released the basic specifications for a "Medical Insurance Imaging Cloud," promoting cross-institutional and cross-regional interoperability of imaging data. As of the end of June, the cumulative index data for the national medical insurance imaging cloud approached 440 million entries. The former opens up application demand, while the latter builds the data foundation. Medical imaging AI is transitioning from single-hospital pilots towards a more systematic phase of infrastructure development.
DIAGENS-B is increasing investment to capitalise on this industrial window. In the first half of the year, the company's R&D costs reached RMB 64.118 million, a 67.4% year-on-year increase, equivalent to approximately 59% of revenue for the period. Among these costs, computing service expenses totalled RMB 45.792 million, accounting for about 71.4% of R&D spending, primarily directed towards base model upgrades, data governance, professional workflow construction, unified evaluation, and core product development. The company's loss for the period widened to RMB 55.883 million, mainly due to increased R&D investment, listing-related expenses, higher sales and administrative costs, and reduced other income. From an operational structure perspective, DIAGENS-B currently exhibits a pattern of "high-speed core business growth running parallel with high platform investment," rather than a loss of revenue momentum. As of the period's end, the company held approximately RMB 655 million in cash and cash equivalents, net current assets of about RMB 701 million, and a debt-to-asset ratio of roughly 13%, providing ample financial space for continued R&D, product registration, and commercial expansion.
Conclusion: DIAGENS-B is Selling the Productivity of Medical Imaging AI
The most important message from DIAGENS-B's interim report is not just that model service revenue doubled year-on-year or that it constitutes nearly 90% of total revenue. It is that several key chains for commercialising medical imaging AI have begun operating simultaneously. The base model can continuously incubate specialist tasks. The data platform supports professionalised production. The Class III certificate validates the ability to transform into a regulated product. Multiple service models provide real revenue channels for model capabilities. As a result, DIAGENS-B's identity becomes clearer. It is not just a medical large model company, nor just a provider of chromosome diagnostic equipment and software. It is building a platform that accelerates the research, production, and application of medical imaging AI, connecting data, experts, models, regulation, and clinical use. The first half of medical AI addressed whether models could function. The second half must answer whether models can be continuously produced, delivered at scale, and generate revenue. With model service revenue approaching 90% of total income, DIAGENS-B has entered this latter stage first. As more specialist tasks, medical institutions, and regulated products connect to the same platform, what DIAGENS-B outputs will no longer be isolated AI tools but the underlying productivity needed for intelligent medical imaging. This is the long-term value behind the company's interim report that deserves the most attention.