Polish AI company Ingenix tops the cellular aging benchmark and open-sources a version of the model.

Poland-based AI and biotech firm Ingenix said its Thalamus Reasoning model ranked first among 29 models on LongevityBench, an independent benchmark for language models used in cellular aging research, and released an open-source version called Thalamus Decision for academic and commercial use. The 9-billion-parameter model, built by a small team, averaged a rank of 1.42 across 14 ranking tasks versus 4.42 for the next-best model and 9.85 for Claude Opus 4.6, and placed first on 11 tasks; it reported a DNA methylation age error of 3.36 years versus 12.9 years for the best text-prompt model on the same Qwen3.5-9B base and 4.67 for the Horvath clock, plus transcriptome ranking accuracy of 0.828 versus 0.671 for the next-best model. Ingenix said the model converts methylation, gene expression and proteomic data into mathematical vectors readable by the model, a process it calls Modality Fusion, using 4 tokens per data record, and that training did not use benchmark data, avoiding contamination. The company, founded in Warsaw in 2023 by the team behind Snowflake-acquired Applica and backed by a €13 million round led by Sofinnova Partners, is developing a 950-billion-parameter version. The result matters as a test of whether reasoning models built around raw biological measurements can outperform frontier labs and specialized models on independent benchmarks. Next, watch adoption and reproduction of the open-source Thalamus Decision and progress on the larger model.