MoonshotAI: Kimi Latest

MoonshotAI: Kimi Latest is a text generation model from Moonshot AI, published under the model id ~moonshotai/kimi-latest. As the provider description states, this model always redirects to the latest model in the Kimi family, so a request to this id resolves to whichever Kimi model is currently the newest rather than to a frozen snapshot. That makes it a convenient choice if you want to stay on the current Kimi generation without tracking version names yourself.

The model accepts text, image and video as input and returns text. Its context window is 1,048,576 tokens, and the maximum output is 943,718 tokens. The large context is the headline capability here: it lets you pass very long documents, long transcripts or extended multimodal material in a single request, and still leave room for a substantial generated response. Supported parameters include temperature, top_p, top_k, min_p, frequency_penalty, presence_penalty, repetition_penalty, max_tokens, stop, seed, logit_bias, logprobs and top_logprobs. It also supports response_format and structured_outputs for constrained output, tools and tool_choice for function calling, and reasoning controls through reasoning, reasoning_effort and include_reasoning.

Those capabilities point to a few practical uses. Long-context work such as summarising or querying large document sets, codebases or archives fits naturally, since you can supply the material directly instead of building a retrieval pipeline first. Because images and video are accepted as input, the model can also handle tasks that mix media with text, for example describing or extracting information from visual material alongside written instructions. Tool calling and structured outputs suit agent-style workflows and any pipeline that needs machine-readable responses, while the reasoning parameters let you trade off between quick answers and more deliberate ones.

You call it through api14's OpenAI-compatible API using the model id "~moonshotai/kimi-latest". Point your existing OpenAI client at the api14 base URL, send a chat completion request with that id, and pass any of the supported parameters listed above in the request body. Because the id redirects to the latest Kimi model, the underlying model behind the id can change over time, so if you depend on exact reproducibility, pin behaviour through the parameters you control, such as seed and temperature, and test after changes.

One thing to keep in mind is that the redirect behaviour is a trade-off. It keeps you current, but it also means the model serving your requests is not fixed, which can matter for evaluation and regression testing. Output is text only, so tasks that require generating images, audio or video are out of scope here. If your workload is small and short-context, the very large context window and output limit may be more than you need, and a smaller model could be simpler to reason about.