Kimi 2.6 and Kimi Code review vs Claude Code

The speaker reviews Kimi 2.6 through Kimi Code, Moonshot AI's CLI coding assistant, using a luxury car hire website as the test project. He reports 42-43 tokens per second during the project. These figures describe his session rather than a general performance guarantee. The workflow uses a subscription and the provider's API; the video does not demonstrate running the model on the user's hardware.

The website preview includes Italian and English switching, car hire pages, booking links and footer navigation. The speaker likes the design and checks several interactions, but also encounters a language display issue and a client-side error. Coding is still underway, and he attributes the errors to the unfinished build without establishing their cause. He later says the work took about 15-20 minutes.

His final rating is 8/10, with Opus 4.7 in Claude Code set at 10/10 on his personal scale. He expects Opus to produce more detailed designs, but this is not a controlled side-by-side benchmark. The video also mentions Gemini 3 Flash, GLM 5.1 and MiniMax, with further model tests planned. A separate Harbor SEO sponsor segment demonstrates content generation and site health features.