Ingenix releases Thalamus Decision, an open-source AI model
The startup says its model topped 28 others on LongevityBench, but independent verification was not immediately known.
Ingenix said Monday that Thalamus Reasoning placed first out of 29 models tested on LongevityBench, which measures performance in cellular aging research. The company is releasing an open-source version named Thalamus Decision.
The Warsaw startup said a small team built the 9-billion-parameter model, which beat 28 other published models, among them systems from top frontier labs. Ingenix said Monday's release is intended to help researchers in academia and industry replicate the findings and extend the work.
"Ranking ahead of frontier labs and specialised models on an independent benchmark validates our approach of Modality Fusion," Ingenix CEO Piotr Surma said in a statement.
It was not immediately known whether the benchmark's maintainers have verified or published the ranking. The announcement did not cite a technical paper or explain the methods behind the results.
Ingenix said Thalamus Reasoning took first place in 11 of 14 benchmark tasks and exceeded the strongest non-language-model baseline on eight.
The model's mean rank was 1.42, while the runner-up scored 4.42, according to Ingenix. The company put Claude Opus 4.6, the leading frontier model, at 9.85.
Ingenix said Thalamus Reasoning recorded a DNA methylation age error of 3.36 years. The strongest text-prompted model using the same Qwen3.5-9B base recorded 12.9 years, according to the company. Ingenix said its result also surpassed the Horvath clock's 4.67 years.
For transcriptomics ordering accuracy, Ingenix gave Thalamus Reasoning a score of 0.828, versus 0.671 for the strongest other large language model.
Ingenix said it used no benchmark data in training Thalamus Reasoning, preventing data contamination. The company described LongevityBench as an independent measure of language-model performance in cellular aging research.
Insilico Medicine said in September 2026 that it had open-sourced an AI longevity toolkit with collaborators, including LongevityBench. Insilico Medicine describes itself as a clinical-stage generative AI drug discovery company.
Rather than treating biological information as text, Thalamus Reasoning uses Modality Fusion to convert complex textual data on methylation and expression into mathematical vectors. The process handles proteomics data, too. Each profile takes four tokens, whereas a standard large "language" model such as ChatGPT uses around 1.3 tokens per word, Ingenix said.
"It shows that when a reasoning model reads biological data as it was measured, rather than as text written about it, meaningful results can be delivered," Surma said.
Thalamus Reasoning belongs to the Ingenix Biological Reasoning Engine, which the company says is designed for translational and clinical research and development decisions. Ingenix said focused teams with AI and biology expertise are more likely than hyperscalers tackling every problem at once to produce a step change in AI for drug discovery.
Ingenix was founded in 2023 and is led by researchers and engineers from the founding team of Applica, an AI company acquired by Snowflake in 2022.
The company is also developing a 95-billion-parameter model of its own.
