Moonshot AI curbs cyber-offensive capabilities in new Kimi K3.1 Pro model

Moonshot AI has introduced Kimi K3.1 Pro while simultaneously reducing its cyber-offensive capabilities. The company removed cyber-related data from its reinforcement-learning environments and training corpus to achieve this.
Kimi K3.1 Pro is a 3-trillion-parameter-class open-weight model with a 1-million-token context window. Moonshot plans to publish the model weights by 4 November 2026.
The firm stated, "We aim not to release open weights that would extend the absolute frontier of cyber capability." Internal benchmarks show the model scored 81.0% on cyber tests, compared to 90.5% for a similar version lacking risk-reduction measures.
These results have not been independently verified. Moonshot AI also released K3.1 Pro Highspeed, which operates at two to three times the speed of the standard version.




















































