Moonshot AI Releases Trillion-Parameter Kimi-K2 Model
The new Mixture-of-Experts model brings massive scale to the open-weights community, focusing on complex reasoning and coding tasks with a 128K context window.
Moonshot AI has released the weights for Kimi-K2-Instruct, a new large language model built at a massive scale. The model is described as having a trillion parameters, placing it among the largest open-access models available to researchers and developers.
Built on a Mixture-of-Experts (MoE) architecture, Kimi-K2 is designed for efficiency at scale. The model features a 128,000-token context window, enabling it to process and reason over extensive documents and complex codebases. According to the developers, the model's strengths lie in its agentic capabilities and coding performance, suggesting an aptitude for multi-step, autonomous tasks.
A New Tool for Complex Tasks
The release of a trillion-parameter model is a significant event for the open-source AI landscape. It provides a powerful foundation for building sophisticated applications that require deep reasoning and the ability to maintain context over long interactions. Key features of Kimi-K2-Instruct include:
- Massive Scale: One trillion parameters.
- Efficient Architecture: Mixture-of-Experts (MoE).
- Large Context: 128,000-token window.
- Specialization: Strong performance in coding and agentic reasoning.
The model weights and card are now available for download from the Hugging Face Hub. The model is released under a custom license, and developers should review its specific terms of use before integrating it into their work.
Sources
- Visit
moonshotai/Kimi-K2-Instruct
Hugging Face
More from Moonshot AI
All Moonshot AI releases →
Moonshot AI releases Kimi K3, a 2.8T-parameter MoE model
The open-weights multimodal model leans into coding and agentic tasks, extending Moonshot's Kimi line into a new scale bracket.
Moonshot AI Releases Kimi, a Multimodal Coding Model
The new Mixture-of-Experts model from the Chinese AI company can generate code while also understanding visual inputs, a rare combination in open models.
Kimi K2.6 tops closed models in coding test
Moonshot AI's open-weights mixture-of-experts model reportedly outperformed Claude, GPT-5.5, and Gemini on a programming challenge.
More in Vision-Language
All Vision-Language →
Mistral Large 4 Arrives as a 1T-Parameter MoE
The new flagship pairs a trillion total parameters with 52B active per token and adds vision alongside multilingual text.
Baidu releases Unlimited-OCR 2 under MIT license
The multilingual vision-language model targets document text recognition and ships with a permissive license for commercial use.
Liquid AI's LFM3-Omni-2B brings any-to-any AI to the edge
The 2-billion-parameter model combines vision, audio and speech-to-speech in a package small enough to run on device.
0 comments
No comments yet. Be the first to weigh in.