Moonshot AI is redirecting its Kimi language models away from coding tasks toward biology, chemistry, and pharmaceutical research. The Beijing startup, DP Technology, stated at the Web Summit in Lisbon that science requires more advanced AI capabilities than law or finance. Giacomo Ficari, the international director, highlighted this strategic shift as benchmark results in software development show diminishing returns.
The company is now integrating Kimi K3.1 Pro into BioMaster, a research assistant designed to handle complex biological data. In a specific test, the system analysed gene data from 3,696 mouse pancreatic cells and identified 100 candidate genes in 90 minutes, a task estimated to take humans four hours. However, independent mathematical breakthroughs remain distant goals as current models struggle to solve open conjectures without human intervention. Industry adoption also lags behind initial optimism. A June survey by Citi found that while 72 percent of US research leaders use AI broadly, only 24 percent expect it to noticeably improve drug development success rates.
* Kimi K3.1 Pro flagged candidate genes in 90 minutes versus an estimated four hours for human researchers.
* Leading models failed to crack any open mathematical conjectures on the TCSExam benchmark independently.
* Only one quarter of surveyed US research leaders believe AI will significantly enhance drug development outcomes.




