Moonshot AI unveiled Kimi K3.1 Pro on Oct. 23, 2026, saying it improves programming, financial analysis and 3D modeling. The model is available through Moonshot’s services, but the company only expects to publish its full weights by Nov. 4, 2026. Moonshot says it has a 1 million-token context window and 2.8 trillion total parameters, with 104 billion active.
K3.1 Pro scored 65.7% on Terminal-Bench 4.0, above GPT-6 Astra’s 58.2% and below Claude Opus 5.5’s 66.4%. Moonshot reported that hallucinated statements fell from 23.4% for K3 to 8.3% for K3.1 Pro on an internal test designed to provoke errors; it cautioned that this is not the model’s overall error rate in ordinary use. Moonshot also said it removed cybersecurity-related training data while preparing the open version; on an internal security test, the final model completed fewer tasks than an unrestricted control version.
