Model Positioning and Capabilities
kimi-thinking-preview is a preview version of a long-thinking model released by Moonshot AI. The official description presents it as a model with multimodal and general reasoning capabilities that excels at deep reasoning and can be used to handle complex coding, mathematics, and work problems. The announcement did not provide a detailed list of all supported modalities or their interface specifications. (platform.kimi.com)
Usage and Limitations
- API responses use
reasoning_contentto return reasoning content andcontentto return the answer; in streaming output, reasoning content appears before the answer. - Reasoning content and the answer are jointly subject to the
max_tokenslimit. In multi-turn conversations, there is no need to put reasoning content back into the context. - At release, the preview version did not support tool calling, online search, JSON Mode, Partial mode, or Context Caching. The official recommendation is to use streaming output and set
max_tokensto at least 4096. (platform.kimi.com)
Release and Service Status
- Release date: 2025-05-06. (platform.kimi.com)
- Current status: As of 2026-11-18, the official model list identifies the model as taken offline; the list does not specify the date it was taken offline. (platform.kimi.com)
Pricing
As of 2026-11-18, the official domestic and international pricing pages had been consulted, but neither the RMB nor USD input and output unit prices for this model could be verified, so no amounts are listed. The official platform charges based on input and output token usage; prices for other models cannot be applied to this model. (platform.kimi.com)
Official Sources
Date checked: 2026-11-18.