All news

X user posts screenshots of instructions for a Kimi K3.1 Pro ad campaign telling creators not to use the paid-partnership label; authenticity and authorship unconfirmed

Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .

The launch of an open model is accompanied by a dispute over hidden advertising; for now it is an unproven allegation, while the outlet confirms the model's figures separately.

According to VentureBeat, X user Elshayib posted screenshots of what he called instructions for a Kimi promotional campaign, which say "do not put on the paid partnership label." The document's authenticity has not been independently confirmed, and it is unclear whether it comes from Moonshot, an outside agency or another party; the outlet found no confirmed response from Moonshot at the time of publication. VentureBeat separately reports the model's results: Terminal-Bench 4.0, 67% (Claude Sonnet 5.5, 64%); CyberGym-E2E-AA, 95%; GDPval-AA, 1,846 Elo (third place); on scientific tasks the model trails GPT-6 Astra by 30 points in one test.

History

Kimi K3.1 Pro: release and dispute over undisclosed promotion

2 events
  1. X user posts screenshots of instructions for a Kimi K3.1 Pro ad campaign telling creators not to use the paid-partnership label; authenticity and authorship unconfirmedOpen now
  2. Kimi releases K3.1 Pro: an open model in the 3-trillion-parameter class, ranked fourth on the independent Artificial Analysis index, with weights promised for November 4
Follow on Telegram