Infinia Technologies Unveils LLM Nia 1.0

Infinia Technologies today introduced Nia 1.0, a healthcare focused language model developed as a research and engineering initiative and evaluated against a broad range of published medical benchmarks. Further, Infinia is opting to extend access to qualified organisations through a controlled evaluation process rather than releasing the model openly.

Nia is a large language model developed by Infinia AI research and engineering teams using a combination of multiple AI techniques, models, and proprietary optimisation methods. It was not trained from scratch as a single standalone model, nor was it simply post-trained from one base model. Instead, its development involved integrating and refining multiple approaches to achieve the desired performance, reliability, and domain-specific capabilities. Certain elements of its architecture, training methodology, and optimisation processes are, naturally, proprietary.

Published benchmarks evaluated independently of any specific deployment indicate that Nia records leading results among comparable same class medical models. Weighed against MedGemma 27B, its closest peer of similar size, Nia leads on four of five text benchmarks, including expert level clinical reasoning and medical licensing examination questions. Full methodology and results, including the one benchmark on which Nia trails, have been published at nia.iamsaif.ai.

“Nia was built on a simple premise. Healthcare AI has to earn trust before it earns scale. We chose controlled access over open release, and we chose to prove the model inside our own healthcare operation first, because the decisions these systems support carry real weight. Publishing our full benchmark results, including the one we do not lead, is part of the same commitment,” said Subhransu Sekhar Mahapatra, CTO, Infinia Technologies.

Having established this performance through broad research and evaluation, Infinia is now validating Nia’s practical application in a live production environment. Apeiro Digital, Infinia’s Kenya healthcare arm has deployed the model to support an e-claims adjudication workflow, supporting the administrative processing of healthcare claims. Its team of medical experts and doctors provides the clinical guidance and domain expertise that informs the engineering and AI development work carried out by Infinia, helping ensure that solutions are aligned with real-world medical needs and standards.

“Running Nia inside a live claims workflow, rather than a controlled pilot, is what tells us where the model genuinely helps and where it does not. That is the operating standard we hold across the Group: prove a system inside a real environment, keep it inside the decisions it is qualified to support, and let the results, not the announcement, determine what comes next,” said Shiraz Sonday, COO, Apeiro Digital.

Controlled Access, Not Open Release

Nia’s benchmark results reflect its broader development and evaluation programme, rather than any single deployment. Kenya’s implementation is best understood as a real-world validation of a model already proven on published benchmarks, not the basis for those results. It is not used for clinical diagnosis or treatment decisions. Infinia is using its own healthcare operation as the first production environment for Nia before extending access to external organisations, applying the same controlled evaluation standard it intends to require of every future deployment.

Nia has not received regulatory approval or clinical certification, and Infinia is positioning the model accordingly. Its current use supports operational and administrative workflows such as claims processing rather than direct clinical decision making, and access is being extended deliberately so that use in sensitive settings can be evaluated with care.

Infinia is taking a deliberate approach to how Nia becomes available. Rather than releasing the model openly, it is extending access to qualified organisations through a structured evaluation process, so that use in healthcare settings can be assessed case by case.

Organisations interested in evaluating Nia can access it through the Nia playground within the Infinia Token Factory environment or apply for access at nia.iamsaif.ai. Infinia expects to extend the evaluation programme to additional qualified organisations in the coming months, building on this initial deployment.