Moonshot AI launches new flagship Kim K3.1 Pro model with 3T parameters
Moonshot AI today introduced Kim K3.1 Pro, a large language model that can perform some tasks better than Claude Opus 5.5 and GPT-6 Astra. The LLM is a post-trained, or upgraded, version of an algorithm that the Chinese artificial intelligence company released in July. The latter model, Kimi K3, comprises 896 individual neural networks with […] The post Moonshot AI launches new flagship Kim K3.1…
Moonshot AI unveiled Kim K3.1 Pro, a new flagship large language model that outperforms Claude Opus 5.5 and GPT-6 Astra on certain tasks. This model is an enhanced version of the previously released Kimi K3, which boasts 896 neural networks and 3 trillion parameters. To keep costs down, Moonshot AI activates only 16 neural networks at a time.
The model's architecture features Kimi Delta Attention, a technology that reduces RAM usage during prompt interpretation. Moonshot AI's post-training process required more processing power and ran on an upgraded stack, but automated quality checks and improved infrastructure maintained efficiency. Kim K3.1 Pro excels in 3D modeling, scoring 92.7% on BenchCAD, which is only 4 percentage points behind Astra and Claude Opus 5.5.
The model can generate both static and animated 3D models, including complex video games. Moonshot AI intentionally excluded cybersecurity data to minimize safety risks, but future models will feature advanced vulnerability analysis. Kim K3.1 Pro will be available through Moonshot AI's paid cloud services, with the model's weights to be open-sourced by Nov. 4.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
