Insilico Medicine Opens ‘Science MMAI Gym’ to Turn General LLMs into Drug Discovery Engines

New updates have been reported about Insilico Medicine.

Insilico Medicine (HKEX: 3696) has launched Science MMAI Gym, a membership-style training platform designed to convert general-purpose large language models into domain-optimized systems for pharmaceutical R&D, potentially creating a new revenue stream and strengthening the company’s positioning in AI-enabled drug discovery. Leveraging more than a decade of internal AI development, a clinical-stage pipeline that includes over 27 preclinical candidates and multiple Phase I and IIa trials, and millions of proprietary chemistry and biology data points, Insilico will allow pharma, biotech, AI labs, and cloud providers to bring their own base models for customized optimization along three tracks: Chemical Superintelligence (CSI), Biology/Clinical Superintelligence (BSI), and a longer-horizon Pharmaceutical Superintelligence (PSI). The Science MMAI Gym architecture combines multi-task supervised fine-tuning and reinforcement fine-tuning with reward models, benchmarked against public and internal datasets such as TDC, TargetBench, and ClinBench, with internal case studies showing up to 10x performance gains on key drug discovery tasks and substantial improvements in ADMET prediction, target identification, and Phase 2 outcome forecasting compared with baseline frontier LLMs that often fail 70–95% of such tasks without domain training.

The program is structured as flexible memberships ranging from intensive two-week sprints to extended PSI engagements, with partners receiving their own CSI/BSI/PSI-enhanced model, detailed benchmark reporting, and optional wet-lab validation via Insilico’s automated assay platforms. By commercializing its AI training stack rather than limiting it to internal pipeline work, Insilico is effectively productizing its “AI gym” as infrastructure for the broader life sciences and AI ecosystem, which could diversify revenue beyond milestone and licensing income from its therapeutics portfolio. Management positions the Science MMAI Gym as a core step toward its Pharmaceutical Superintelligence roadmap, aligning with the company’s strategy to integrate AI and automation across drug discovery and adjacent verticals such as advanced materials and agriculture. Coming shortly after its listing on the Main Board of the Hong Kong Stock Exchange in December 2025, the launch underscores Insilico’s effort to monetize its data assets and models at scale while deepening integration with external R&D pipelines and cloud platforms, potentially increasing switching costs and strengthening its competitive moat in AI-driven biotech.