INFevo (无界进化), a Beijing company building AI models of cells for drug research, said on 16 November 2026 that it had raised more than RMB 100m (about $15m, according to KrASIA) in a Pre-A round.
New investors Star Chain Capital (星连资本), Prosperity7 Ventures, Huagai Capital, Huakong Fund and DT Capital took part, and existing shareholders Shunwei Capital, HongShan and Songhe Capital invested again. No lead investor or valuation was disclosed.
The round came five months after an angel round of tens of millions of yuan in June 2026. Star Chain Capital, which announced the deal, described it as the largest single financing so far for a Chinese virtual-cell company.
What INFevo does
INFevo builds so-called virtual-cell models, which try to predict how cells will change when exposed to a drug or a genetic change before any lab work is done. The aim is to help researchers rank candidate molecules, compare drug targets and spend scarce lab time on the most promising ideas.
Its own foundation model family is called OCOO. The first released version, OCOO-T, models changes in gene expression from transcriptome data. The company reported leading results on three public single-cell perturbation benchmarks, Tahoe100M, Replogle and PBMC, which cover chemical, genetic and cytokine perturbations. Unite.AI noted that these are developer-reported results.
The models sit alongside an automated wet lab that handles cell culture, perturbation and multi-omics measurement. Predictions help choose experiments, and the lab results are used to check the predictions and improve the model. INFevo calls the wider approach recursive scientific discovery: questions, hypotheses, methods and results, including failed experiments, are stored and linked so they can guide later work.
Research agents and clinical work
The company has also launched The Popper Project, a tool named after the philosopher Karl Popper that uses research agents and databases to generate, test and refine hypotheses with traceable reasoning. Its Karl Workbench research tool is in closed beta. INFevo said that in an October retest on the SciAgentArena benchmark, Karl Workbench running on a DeepSeek model ranked first overall in single-cell and spatial omics tasks. Unite.AI pointed out that the result covers those two categories only.
INFevo is also working with a Beijing key laboratory led by Peking University First Hospital on organoids for liver, biliary and pancreatic tumours, where it supplies the AI and virtual-cell work and tests its predictions on patient-derived samples.
Backed by XtalPi
INFevo was incubated inside XtalPi (2228.HK), the Hong Kong-listed AI drug discovery company. XtalPi co-founder and chief innovation officer Lai Lipeng is also an INFevo co-founder. The two companies say XtalPi's molecule design and automated synthesis could be combined with INFevo's cell-response predictions to link molecule generation, prediction and lab testing.
Co-founder Zhao Yalong said the round would be used to build out the combined model and automated-lab infrastructure and speed up work on its recursive discovery system.
What the money is for
INFevo said the funds will go mainly to foundation model development, expanding its high-throughput automated lab, and deeper partnerships with drug companies, hospitals and research institutions. The company did not disclose its headcount.
