Insilico opens AI longevity toolkit for researchers worldwide

A test bed for aging biology

Insilico Medicine has released an open research package built to test and deploy artificial intelligence in aging biology, alongside a cover study in the September 17 issue of Cell. The package combines LongevityBench, specialized open-source language models and an agentic platform called Longevity Claw.insilico The company says the work involved collaborators from Liquid AI, the Buck Institute for Research on Aging, Harvard Medical School and Brigham and Women’s Hospital.insilico

The release targets a stubborn problem: aging research spans clinical records, genetics, epigenetics, transcriptomics and proteomics, yet general-purpose models can perform unevenly across those data types. That challenge mirrors the broader field’s effort to establish biomarkers that can reliably identify and evaluate longevity interventions.cell Earlier research has also argued that pairing AI with open science could improve the discovery of disease targets and therapeutic candidates.cell

Smaller models challenge frontier systems

LongevityBench was designed to reward analysis of unfamiliar biological measurements rather than recall of scientific facts. Insilico reported that no frontier model led across all five tested biological domains and that performance shifted when questions were rephrased. Predicting biological age directly from omics measurements proved the hardest task.insilico

The researchers then trained five domain-specific models ranging from 0.6 billion to 9 billion parameters. The top model, L-Qwen3.5-9B, recorded the best overall result among 26 systems assessed, while even the smallest specialized model beat most of the frontier systems in the company’s evaluation.insilico Those results suggest curated scientific data and targeted fine-tuning can matter more than raw scale for narrow biological tasks, although an open benchmark will now let outside researchers test that conclusion.

Target discovery still needs laboratory proof

Longevity Claw links the strongest specialized model to tools for aging-clock calculation, gene-set enrichment, evidence retrieval and target prioritization. Applied across 14 hallmarks of aging, it nominated 328 genes and produced enrichment of up to 5.6-fold against an independently published set of experimentally supported aging targets.insilico That is a prioritization signal, not evidence that the proposed interventions will work in animals or people.

The toolkit arrives as commercial interest in AI-guided longevity research is expanding. Eli Lilly agreed to a drug-discovery collaboration with Insilico worth up to $2.75 billion, while the biotech’s pulmonary-fibrosis candidate has separately been studied through six proteomic aging clocks in a Phase IIa trial.medicaldaily +1 The models, training resources, benchmark data and evaluation code are being made available for independent use, shifting the next test from an in-house leaderboard to reproducible results across research groups.insilico