MoonshotAI: Kimi K3.1 Pro

Kimi K3.1 Pro is a text model from Moonshot AI, released by the provider on 2026-10-23. It is described as Moonshot AI's flagship open-weight multimodal reasoning model and a post-trained successor to Kimi K3. The model id on api14 is moonshotai/kimi-k3.1-pro, and it sits in the text category.

The model accepts text, image and video as inputs and produces text as output, so it can work over mixed media prompts rather than text alone. Its context window is 1,048,576 tokens, which is roughly one million tokens, and the maximum output is 943,718 tokens. That combination of a very large input window and a very large output budget is the defining practical characteristic of the model.

Those capabilities point to work that involves long inputs and long answers. Long-horizon coding tasks fit, since a large repository or a long specification can be held in context while the model produces substantial code in return. The same applies to reasoning over lengthy documents, and to tasks where images or video are part of the input and the answer is written text, such as describing or analysing visual material alongside a long written brief. Because the output limit is high, the model can be asked for extended responses rather than short summaries.

You call it through api14's OpenAI-compatible API using the model id "moonshotai/kimi-k3.1-pro". The request parameters it supports are frequency_penalty, include_reasoning, max_tokens, presence_penalty, reasoning, reasoning_effort, response_format, stop, structured_outputs, tool_choice and tools. The reasoning and include_reasoning parameters, together with reasoning_effort, let you control how reasoning is requested and returned, while response_format and structured_outputs are available when you need the output in a defined shape. The tools and tool_choice parameters support function calling, so the model can be used inside agent or tool-using workflows.

There are no stated restrictions in the facts beyond the model being text output only, so if you need image or video generation this model is not the fit. If your task is short and purely textual, the very large context and output limits may be more than you need, and a smaller text model could be the simpler choice. For long-context, multimodal-input, reasoning-heavy or tool-using work, this model is the appropriate option among those described here.