A LinkedIn post from Insilico Medicine highlights results from Day 24 of its ScienceAIBench series, focusing on AI models for ophthalmic drug target discovery. The benchmark evaluates models’ ability to identify clinical-stage targets across multiple eye diseases using Mean Average Precision at TopK.
Claim 55% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
According to the post, GPT-5 led performance in complex retinal and systemic indications, including macular degeneration, diabetic retinopathy, and thyroid-associated ophthalmopathy. DeepSeek R1 appeared strongest in hereditary conditions such as corneal dystrophy and retinitis pigmentosa, while Gemini 2.5 Pro and Grok 4 led in glaucoma and pterygium, respectively.
The post also points to a notable gap where GPT-5 scored 0.000 on corneal dystrophy while DeepSeek R1 achieved a perfect 1.000, suggesting model-specific blind spots tied to genetic dystrophies. For investors, these differentiated performance profiles may underscore both the technical sophistication of Insilico Medicine’s benchmarking platform and the importance of model selection in high-stakes, tissue-specific drug discovery.
This type of comparative evaluation could strengthen Insilico Medicine’s positioning as an independent validator of AI systems for biotechnology applications. If the ScienceAIBench series gains traction with pharma and biotech partners, it may create opportunities for platform licensing, collaborations, or data-driven services that support long-term revenue and enhance the company’s strategic role in AI-enabled drug development.