# Moonshot AI Officially Releases Kimi K3.1 Pro Weights

- Category: Flash (快讯)
- Published: 2026-11-05T11:54:00.000Z
- Source date: 2026-11-05
- HTML: https://www.88z.ai/en/news/flash-20261105-4a908a434e4bad59
- Markdown: https://www.88z.ai/en/news/flash-20261105-4a908a434e4bad59.md
- Other language: https://www.88z.ai/news/flash-20261105-4a908a434e4bad59.md

Moonshot AI has officially open-sourced the weights for Kimi K3.1 Pro on Hugging Face. Featuring 2.8 trillion parameters and a mixture-of-experts architecture, the model supports context windows of up to one million tokens. It ranks first among open-weight models in agent benchmarks and leads in Bash reliability compared to similar open-source counterparts.

## At a glance

- Kimi K3.1 Pro weights are now open-source on Hugging Face, supporting one-million-token contexts.
- The model ranks first among open-weight models in Agent Arena and leads in Bash reliability.

## The update

Moonshot AI has officially released the model weights for Kimi K3.1 Pro on the Hugging Face community, marking the full opening of this high-performance large language model to developers. Kimi K3.1 Pro is a mixture-of-experts (MoE) model with a scale of 2.8 trillion parameters. It activates approximately 104 billion parameters per inference and supports long context windows of around one million tokens. In terms of performance evaluation, data from the Agent Arena leaderboard, based on over 17,500 real-world agent sessions, shows that Kimi K3.1 Pro ranks first among open-weight models and 13th overall. The model achieved a net improvement rate of 5.1%. Notably, it leads all open models in metrics for Bash command execution reliability and steerability, demonstrating robust capabilities in handling agent-based tasks.

## Sources

- [IT之家](https://www.ithome.com/1/022/490.htm) (2026-11-05)

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.