Moonshot AI launches new flagship Kim K3.1 Pro model with 3T parameters

AI

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 3 trillion parameters. It only activates 16 of those neural networks at any given time to minimize inference costs.

Moonshot AI also optimizes its models’ hardware footprint in other ways. Kim K3 features Kimi Delta Attention, a custom implementation of a memory-saving technology called Gated DeltaNet. It reduces the amount of RAM that the model’s attention mechanism uses to interpret user prompts. The feature was presumably carried over to Kim K3.1 Pro.

The company says the new model’s post-training workflow consumed significantly more processing power than similar projects it completed in the past. Additionally, it ran on an upgraded version of the company’s post-training stack. Moonshot AI added automated quality checks and boosted the underlying infrastructure’s efficiency.

One area where Kimi K3.1 Pro delivers particularly competitive output quality is 3D modeling. There are two ways to create a 3D model of an object: engineers can draw it or they can write a code snippet that specifies its geometry in numerical terms. Kimi K3.1 Pro takes the latter approach. The LLM scored 92.7% on a 3D model coding benchmark called BenchCAD, which put it within four percentage points of Astra and Claude Opus 5.5.

Kimi K3.1 Pro can not only generate static 3D models but also animate them. In one test, Moonshot AI used the model to generate an animated simulation of human motion. Another internal evaluation saw the company create multiple 3D video games. Games are difficult for AI models to generate because they contain a large amount of code and even a single bug can crash them.

Pro scored 65.7% on the industry standard Terminal-Bench 4.0 coding benchmark, which put it more than 7% ahead of Astra but slightly behind Opus 5.5. Furthermore, it managed to outperform both models on OfficeQA Pro v2. It’s a test that measures LLMs’ ability to perform complex financial analysis tasks.

The model’s performance on popular cybersecurity benchmarks significantly trails the scores of Astra and Opus 5.5. According to Moonshot AI, its engineers deliberately removed cybersecurity-related information from the model’s post-training dataset to address safety risks. The company plans to equip its future models with more advanced vulnerability analysis features.

Kimi K3.1 Pro is available via Moonshot AI’s paid cloud services on launch. The company plans to open-source the model’s weights by Nov. 4.

Image: Unsplash