Insilico Medicine said it has published a study in Cell introducing an open AI toolkit for research into aging and longevity. The toolkit consists of LongevityBench, a family of language models called Longevity-LLMs, and an agentic research platform named Longevity Claw. The paper was the cover feature of the journal's Sept. 17, 2026 issue, and the work was conducted with researchers from Liquid AI, the Buck Institute for Research on Aging, and Harvard Medical School and Brigham and Women's Hospital, the company said.
That publication follows a Sept. 7, 2026 study in Nature Biotechnology, the company said, which reported that rentosertib, its AI-discovered and AI-designed drug candidate for idiopathic pulmonary fibrosis, reduced biological age across six independent proteomic aging clocks in a Phase IIa clinical trial.
LongevityBench was built to test whether AI systems can reason across the data types that describe human aging, spanning five domains: clinical data, genetics, epigenetics, transcriptomics and proteomics. The release said the researchers evaluated 18 frontier AI systems, including models from OpenAI, Google, Anthropic, xAI, DeepSeek and Moonshot AI. No single frontier model was strongest across all five data types, and performance shifted depending on how questions were phrased. Predicting biological age directly from omics measurements was the hardest task, and even the largest frontier models struggled with it.
The team also developed five compact open-source longevity models, ranging from 0.6 billion to 9 billion parameters, fine-tuned on aging-specific clinical and multi-omics data with Insilico's MMAI Gym for Science. The family was built on open architectures from Liquid AI and Alibaba, including LFM2 and the Qwen3 and Qwen3.5 model families. Despite their smaller size, the company said the models matched or exceeded all 16 frontier systems evaluated on LongevityBench. The best performer, L-Qwen3.5-9B, posted the highest overall score among all 26 systems in the study, ahead of Google's Gemini 3.1-Pro, while the smallest model, at roughly 0.6 billion parameters, outperformed most of the frontier systems tested.
The researchers embedded L-Qwen3.5-9B in Longevity Claw, an open-source agentic platform that pairs the model with tools for gene-set enrichment analysis, biological aging-clock calculation, population-level profiling, evidence retrieval and synthesis, and candidate target evaluation and prioritization. The release said the platform was deployed across 14 recognized hallmarks of aging and nominated 328 genes as potential targets for intervention. Against an independently published reference set of experimentally supported aging targets, those candidates showed statistically significant enrichment of up to 5.6-fold. One nominated gene, KDM1A, was independently validated in a separate published study, in which its modulation extended lifespan in C. elegans.
Insilico said it is releasing the benchmark, the specialized models, training resources, evaluation code and the Longevity Claw platform so that researchers can test, validate and further develop them.
