Model Positioning
Kimi-Researcher is an autonomous agent model for complex research tasks, based on an internal Kimi k-series model and trained through end-to-end agent reinforcement learning. The official announcement was published on June 20, 2025. (moonshotai.github.io)
Capabilities and Uses
- It can perform multi-step planning, reasoning, and tool calls using parallel real-time search, a text browser, and code execution tools. The official report states that it performs approximately 23 reasoning steps and explores more than 200 URLs per task on average. (moonshotai.github.io)
- It can iteratively retrieve and cross-check information, synthesize materials, and generate research reports with citations. Use cases include academic research, legal and regulatory reviews, difficult information retrieval, clinical evidence reviews, and corporate financial analysis. (moonshotai.github.io)
- The evaluation results reported officially include Pass@1 of 26.9% and Pass@4 of 40.17% on Humanity's Last Exam, and Pass@1 of 69% on xbench-DeepSearch. These results were obtained at a specific time and under specific evaluation conditions and do not guarantee the same level of performance on actual tasks. (moonshotai.github.io)
- The official help center states that Kimi Deep Research is powered by this model and can be accessed through the web or the Kimi App, producing traceable in-depth reports and visual reports. (kimi.com)
Limitations
Changes to tools such as search engines may affect results; generated content and citations still require human verification, especially in professional settings such as medicine, law, and finance. The official help center states that Deep Research typically takes 10–25 minutes and is not suitable for simple question answering, creative writing, or fixed-template tasks. The announcement describes a context of hundreds of thousands of tokens in long trajectories, but does not explicitly specify the maximum context window available to users. (moonshotai.github.io)
Pricing and Access
As of November 18, 2026, a review of the official domestic and international model pricing pages did not verify any RMB or USD API input or output pricing specifically for Kimi-Researcher, nor did it verify a public API ID for the model; rates for other Kimi models are not applied to this model. The Kimi Deep Research product uses a unified quota, and its product quota is not equivalent to an API token price for this model. (platform.kimi.com)