Moonshot AI shifts Kimi models from coding to scientific research

Covered by 2 sources
Read onSouth China Morning Post
Moonshot AI is repurposing its Kimi language models to focus on biology, chemistry, and pharmaceuticals, integrating DP Technology to enhance gene analysis. The move aims to advance AI-driven scientific discovery amid ongoing challenges in theoretical science and drug development.

What happened

Moonshot AI, a Beijing-based startup, has redirected its Kimi language models from coding applications to scientific research in areas such as biology, chemistry, and pharmaceuticals. The company is collaborating with DP Technology, which is incorporating Kimi K3.1 Pro into its BioMaster research assistant for gene analysis. This shift comes as progress in coding benchmarks has slowed and as the company seeks new applications for its technology. The move places Moonshot AI in competition with US firms working on AI-driven scientific discovery.

Key details

  • Kimi K3.1 Pro was used to analyze gene data from 3,696 mouse pancreatic cells, identifying 100 candidate genes in 90 minutes.
  • DP Technology is integrating Kimi into its BioMaster research assistant.
  • On the TCSExam benchmark of 100 theoretical computer science problems, leading AI models produced proofs but did not solve any open conjectures independently.
  • A June Citi survey found that 72% of 50 US pharma research leaders use AI broadly, but only 24% expect it to significantly improve drug development success rates.
  • Moonshot AI's international director discussed the shift at the Web Summit in Lisbon.

Why it matters

Researchers in biology, chemistry, and pharmaceuticals may gain faster gene analysis and data processing capabilities through tools like BioMaster powered by Kimi. However, the impact on theoretical science and drug development remains limited, as current AI models have not achieved major independent breakthroughs. Pharmaceutical companies and research leaders face ongoing uncertainty about the practical benefits of AI in improving drug development outcomes.

Also covered by 1 source


More in Machine Learning