Moonshot AI says it is steering its Kimi models toward science as coding gains flatten, SCMP reports

Moonshot steers Kimi from coding toward scientific research

Moonshot AI is shifting Kimi toward biology, chemistry and drug research. In one test, Kimi K3.1 Pro flagged 100 candidate genes from 3,696 mouse pancreatic cells in 90 minutes.

By
Nathan Brooks
Filed
Read
3 min read

Moonshot AI released Kimi K3.1 Pro on Oct. 23, a post-trained version of its 2.8-trillion-parameter K3 model, as the Beijing startup steers Kimi toward scientific research. The South China Morning Post reported the shift from Web Summit in Lisbon.

The Post reported that Moonshot is moving Kimi from coding toward biology, chemistry and pharmaceutical research because the company sees diminishing returns from further coding gains. Moonshot has published neither a blog post nor benchmark results documenting the pivot. The Post's account, summarized by The Decoder, remains a reported strategy rather than a measured outcome.

Giacomo Ficari, Moonshot's international director, said at Web Summit in Lisbon that "science requires smarter AI models than law, finance or code." That is a demanding position for a model whose company has published no benchmarks for the shift.

What has the science test shown?

DP Technology, a Beijing AI-for-science company, is building Kimi K3.1 Pro into BioMaster, its research assistant. In one test, the system analyzed gene data from 3,696 mouse pancreatic cells and flagged 100 candidate genes in 90 minutes. The reports estimated about four hours for a person to do the work.

That result measures one defined task, not general laboratory accuracy. Moonshot has published no benchmark for the pivot. The cited reports provide no independent validation of the screening outcome.

What do the coding scores show?

Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, placing it fourth behind Anthropic's Claude Opus 5.5 at 58. On the Terminal-Bench 4.0 coding test, Moonshot reported 65.7%, compared with 66.4% for Opus 5.5.

On both cited measures, Opus 5.5 scores slightly higher. Neither result evaluates biology, chemistry or pharmaceutical work, so the coding standing says little about the science case. The coding gap is small on Terminal-Bench, while the science case remains unmeasured.

What does the scientific workbench add?

RuntimeWire reported on Nov. 11 that Kimi's desktop app contains an unannounced scientific workbench. The application includes viewers for molecules, genomes and medical scans. RuntimeWire framed the feature as Kimi's answer to Anthropic's Claude Science.

Moonshot has not commented on that report. The app suggests a possible delivery format for the research shift, but the source offers no release date, pricing or laboratory customer list. Science tools are already a competitive field.

Can Moonshot secure the compute?

Compute access is not settled. Bloomberg reported on Nov. 10 that Moonshot rented 32 Nvidia B300 servers housed in China, a claim Moonshot disputes. CNBC reported earlier that month that the company sought access to Nvidia Blackwell clusters in Europe.

The 32-server count remains disputed, and a request for European access is not a confirmed cluster allocation. The reports do not establish where Moonshot can run large research workloads or on what timetable. No procurement value or European deployment date was disclosed.

What would make the shift commercial?

DP Technology provides an early customer use case, not a disclosed commercial scale. The reports give no subscription price, revenue figure or rollout timetable for Kimi's science offering. Moonshot's shift also has no published benchmark trail.

Moonshot's next disclosures will determine whether the science shift becomes a business: published benchmarks, laboratory adoption, pricing and confirmed compute capacity. Until then, Kimi's laboratory ambitions remain a direction rather than a demonstrated market position. If Moonshot secures European Blackwell capacity and converts the gene-screening test into repeatable, priced deployments, biology could become a durable business for Kimi.

Source: mlq.ai

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

Senior reporter covering people, launches and day-to-day practice in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,008 articles

Related articles

« Previous article