AI system with 17 benchmarks aims to accelerate ageing research

September 17, 2026 at 05:00 AM1 min

This digest was compiled by AI from multiple sources — links to the originals are below.

Researchers unveiled an AI system trained on ageing-biology data to advance longevity research, according to a study published in Cell on September 17. The system includes 17 benchmark tasks to evaluate large language models on ageing-related projects. It outperformed larger commercial models from OpenAI and DeepSeek in many tests.

Key Facts

The AI system was described in a study published in Cell on September 17, 2026.
The system includes 17 benchmark tasks for evaluating large language models on ageing-related projects.
Specialist LLMs trained on ageing-biology data outperformed larger commercial models from OpenAI and DeepSeek in many tests.
Marinka Zitnik, a computer scientist at Harvard Medical School, called the work 'a very important contribution to the longevity and ageing field.'

The AI System

The system introduces large language models trained on ageing-biology data. It includes a suite of 17 tasks that serve as benchmarks to gauge LLM performance on ageing-related projects. An interface brings the models and other ageing-related research tools together with AI assistants called agents.

Benchmark Results

The authors used their benchmarks to evaluate both specialist LLMs and larger commercial models from OpenAI and DeepSeek. In many, although not all, of the tests, the LLMs tailored to ageing research outperformed the large LLMs. Marinka Zitnik, a computer scientist at Harvard Medical School, said the paper 'can inform the development of next-generation AI models.'

Defining Ageing

The work tackles the challenge of training AI tools to understand ageing when scientists have not yet defined the concept themselves. Zitnik noted that for cancer, AI tasks can be crisply defined, but 'ageing is a very different beast. It is hard to define.' Researchers have developed various 'clocks' to measure biological ageing using facial features, protein levels, gene activity, and brain scans.