In the domestic artificial intelligence field, the highly anticipated Kimi is set to launch its next-generation trillion-parameter large model, generating great market anticipation. This news was reported by Kuaiji Technology on January 20. Kimi's development company, Moonshot AI, has recently started recruiting testers to prepare for the upcoming new model.

According to Kimi's announcement, the internal testing plan for the new model's API has already begun, and access is currently only available to users who have completed "enterprise certification." Although Kimi has not explicitly revealed the specific version number of the new model, the industry generally suspects it will be an upgrade to the previously released K2 large model, possibly named K2.1 or K2.5. If it is the latter, it could mean a significant performance improvement.

Looking back at the development of the K2 series, the official release of Kimi K2 in July last year and the K2 Thinking reasoning model launched in November achieved SOTA (state-of-the-art) results in multiple benchmark tests, surpassing GPT-5 in the market. At the same time, the API pricing of the K2 series is also very competitive, with both input and output costs lower than GPT-5, attracting a large number of users' attention.

Notably, Kimi's founder Yang Zhilin stated that the company's current valuation has reached $4.8 billion, with cash reserves exceeding 10 billion yuan, providing strong capital support for further R&D of next-generation large model technology. The new generation model will adopt Kimi Delta Attention linear attention mechanism, which is expected to significantly improve its ability to process short and long texts, while incorporating more aesthetic and value design elements, helping it break away from the homogenized competition.

With the development of AI technology, future large models will not be limited to text processing; multimodal and intelligent agent capabilities will become key areas. The launch of Kimi K2.1/K2.5 is expected to bring new breakthroughs in areas such as AI programming capabilities, meeting the growing market demand.