Kimi K3 launches with 2.8 trillion parameters

Kimi K3 launches with 2.8 trillion parameters

N
News Editor
2026-07-16 23:57:56
A new AI model, Kimi K3, was released on July 16 ahead of the 2026 World Artificial Intelligence Conference and the High-level Meeting on Global AI Governance, according to Xinhua as cited by ChainCatcher. The model has 2.8 trillion parameters and was described as the world’s largest open-source model by parameter count at the time of release. Beijing Moonshot AI, the company behind the model, said Kimi K3 natively supports visual understanding and offers a 1 million-token context window. The company said the model was optimized for complex task settings including software engineering, knowledge work, deep research, and multimodal understanding. It also said evaluation results showed Kimi K3’s overall intelligence level was close to leading closed-source models globally. A company representative added that Moonshot AI used a self-developed underlying model architecture during training and built a full set of scientific training methods in the process.
Kimi K3Moonshot AIopen-source modelartificial intelligencelarge language modelvisual understandingtechnology trends

Kimi K3, a new AI model, was released on July 16 ahead of the 2026 World Artificial Intelligence Conference and the High-level Meeting on Global AI Governance, according to Xinhua as cited by ChainCatcher. The model has 2.8 trillion parameters.

Xinhua said it is currently the world’s largest open-source model by parameter count. Beijing Moonshot AI, the company that released it, said Kimi K3 natively supports visual understanding and comes with a 1 million-token context window. The model was optimized for complex task scenarios including software engineering, knowledge work, deep research, and multimodal understanding, the company said, adding that it improves large models’ ability to handle complex tasks.

According to the report, evaluation results showed Kimi K3’s overall intelligence level was close to leading closed-source models globally. A Moonshot AI representative also said the company used a self-developed underlying model architecture while training the new foundation model and built up a full set of scientific model training methods during that process.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
300

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.