Moonshot AI Releases Kimi K3.1 Pro MoE Model
Moonshot AI launches Kimi K3.1 Pro, a 2.8T total and 104B active MoE model supporting a 1M token context window, with open weights arriving November 4, 2026.

Stock photo for illustration only, not from the actual event
- Moonshot AI announced the release of the Kimi K3.1 Pro MoE model
- Features 2.8 trillion total parameters and 104 billion active parameters
- Supports a massive 1 million token context window
- Open weights scheduled for release on November 4, 2026
Artificial intelligence firm Moonshot AI has officially unveiled its latest advanced model, Kimi K3.1 Pro. Built on a Mixture of Experts (MoE) architecture, this model is engineered to handle massive data processing workloads with exceptional precision, targeting enterprise users and developers who demand high-performance natural language processing tools.
The architecture of Kimi K3.1 Pro scales up to a total of 2.8 trillion parameters, while maintaining 104 billion active parameters during inference. This design approach allows the system to achieve high computational efficiency, balancing processing speed and resource consumption while delivering advanced reasoning capabilities.
A standout capability of Kimi K3.1 Pro is its expansive context window, supporting up to 1 million tokens. This massive capacity enables users to input extensive legal documents, lengthy software repositories, or entire books in a single prompt, allowing the model to analyze and comprehend complex relational data seamlessly.

Stock photo for illustration only, not from the actual event
In benchmark evaluations, Kimi K3.1 Pro achieved a score of 65.7% on Terminal-Bench 4.0, demonstrating strong proficiency in handling coding tasks and terminal command executions. Additionally, Moonshot AI priced the API usage at $2 per million input tokens and $10 per million output tokens.
The introduction of a large-scale MoE model with a 1-million-token context window highlights the AI industry's ongoing shift toward enhanced long-context processing capabilities. Such advancements are crucial for analyzing lengthy technical papers, financial reports, or codebases. Furthermore, releasing open weights enables independent researchers and enterprises to fine-tune the model on their proprietary infrastructure, thereby fostering greater innovation within the AI ecosystem.
Moonshot AI confirmed that the open weights for Kimi K3.1 Pro will be officially made available to the public on November 4, 2026, marking another significant milestone in the large language model landscape.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment