AboutAI ToolsCursorClaude CodeLovableBolt.newAI tool reviewsTutorialsVibe codingAI agentsMobile apps with AIAll tutorialsBlogSponsorWork With MeJoin 285K+ Subscribers

18 November 2026 · 3 min read

Ingenix to Open-Source 9B Thalamus Decision Model for Biology

Polish AI startup Ingenix published benchmarks for Thalamus Decision, a typed decision model for biology, and announced the 9B version will be open-sourced on 18 November 2026.

Ingenix Publishes Benchmarks and Open-Source Plans for Thalamus Decision

On 17 November 2026, Polish AI startup Ingenix published performance benchmarks for Thalamus Decision, which the company describes as the world’s first typed decision model for biology [1]. The 9-billion-parameter version of the model will be open-sourced the following day, 18 November.

What Makes Thalamus Decision Different

Unlike large language models that generate free-text answers, typed decision models produce confidence scores for every possible option. This means a system can decide when to act autonomously and when to flag a decision for human review. Until now, typed decision models have been limited to text and, in some cases, images. Thalamus Decision extends this to biological data: it can read bulk and single-cell RNA-seq expression profiles [1].

The model comes in two sizes, 9B and 27B parameters. It is built on Ingenix’s Thalamus Reasoning model, which recently claimed first place out of 29 models on the LongevityBench benchmark [1].

Benchmarks and Capabilities

Thalamus Decision handles a range of biological tasks, including assigning cell type, judging cell malignancy, identifying a sample’s tissue of origin, predicting cell lines’ drug response, and assessing perturbation response – whether silencing a gene changes expression and in which direction. The latter is a critical early-stage process for genetic medicine IND studies [1].

For drug discovery, the model scored 84–89% on two GDSC cell-line drug-response tasks, approximately 94% on gene-essentiality across tissues, and approximately 95% when identifying a cancer model’s type (cell line, organoid, PDX or patient tumour) [1].

The model’s calibrated confidence scores allow it to automate a significant portion of decisions. At a 5% target error rate, the 9B model automated 62.6% of decisions with a realised error rate of 6.3%. The 27B model automated 66.2% of decisions with a 6.8% realised error rate [1].

Open-Sourcing and Availability

Ingenix confirmed that the 9B version of Thalamus Decision will be open-sourced on 18 November 2026. No details were provided about the licence or the platform for the release. The 27B model remains proprietary at this stage [1].

What It Means for Developers and Builders

For developers working at the intersection of AI and biology, an open-source typed decision model that understands RNA-seq data opens several practical possibilities. The model’s ability to output confidence scores rather than plain text means it can be integrated into decision-support pipelines where knowing when to escalate to a human is essential. The benchmarks suggest it can already handle tasks such as cell type classification and drug response prediction with reasonable accuracy, and the confidence calibration provides a measurable way to set automation thresholds.

The open-source release of the 9B model will allow researchers and startups to experiment with the model’s capabilities, fine-tune it on proprietary datasets, and embed it into larger bioinformatics workflows. The fact that it is built on a reasoning model that performed well on LongevityBench may also make it a candidate for ageing research applications.

About Ingenix

Ingenix was founded in 2023 and is headquartered in Warsaw. The company’s Biological Reasoning Engine is built on a Modality Fusion architecture that integrates models across different biological modalities and scales, then reasons across their representations. Every output includes a reasoning trace with mechanism, evidence, and caveats. The founding team includes AI researchers and engineers from Applica, the AI company acquired by Snowflake in 2022, whose TILT model contributed to the GPT-4 benchmark [1].

Sources

  1. Ingenix Publishes Benchmarks for Thalamus Decision Biological AI Model - AIwire
  • ingenix
  • thalamus
  • biology
  • open-source
  • ai

Researched and drafted with AI from the sources above.

Learn to build with AI

Hands-on tutorials on Cursor, Claude Code, AI agents and more.