On February 18, Beijing Moonshadow Technology Co., Ltd. announced the launch of its latest model, kimi-latest, on the Kimi Open Platform, aimed at providing developers and enterprise users with more powerful and stable AI generation capabilities. Since the public beta of the Kimi Open Platform on January 31, 2024, the moonshot-v1 series models have been the core support for the Kimi Smart Assistant. However, with the rapid iteration of the Kimi Smart Assistant, the demand for model stability and functionality among platform users has gradually diversified, prompting Moonshadow Technology to introduce this new model.
The kimi-latest model will always align with the latest model currently used by the Kimi Smart Assistant and will be updated in sync with product upgrades. This feature ensures that open platform users can experience the outstanding effects brought by the latest model, including experimental features that are not yet stable, while maintaining the stability of the original model. The kimi-latest model supports a context length of 128k and can automatically select 8k, 32k, or 128k models as the billing standard based on the context length. Additionally, this model supports image understanding, automatic context caching, and other functions, with a caching hit token cost of only 1/M Tokens, while manual context caching is not yet supported. Other features remain consistent with the moonshot-v1 series models, including ToolCalls, JSON Mode, Partial Mode, and online search capabilities.
The kimi-latest model is suitable for various scenarios. For developers using large model chat applications such as ChatWise, ChatBox, LobeChat, and OpenCat, kimi-latest is the best choice, providing a chat experience similar to that of the Kimi Smart Assistant. For users requiring intent recognition or structured data extraction, the moonshot-v1 series models are more appropriate. Currently, the kimi-latest only supports calls to the regular version of the kimi model, and the Kimi k1.5 long-thinking model does not support API calls at this time. If there are related needs, users can submit a testing application by filling out the Kimi k1.5 application form. Moonshadow Technology will evaluate the application information and contact users who pass the review via email, providing an API key to access the Kimi k1.5 model.
The launch of the kimi-latest model marks a further development for Moonshadow Technology in the field of AI open platforms. By continuously optimizing and upgrading models, the company is committed to providing developers with more efficient and intelligent tools, promoting the application and innovation of AI technology in more fields.