Kimi K3 Generates a 32-Page Chain of Thought

emollick · x · 2026-07-19

While testing Kimi K3, the author asked it to recommend two relevant poems about the "current state of GenAI." The model generated a chain of thought spanning 32 pages, which the author noted contained numerous loops and dead ends.

The key takeaway here isn't the poetry itself, but K3's reasoning performance on complex, open-ended tasks: the thinking process is incredibly lengthy, but its overall stability is somewhat lacking.

Related event: Testing Kimi K3: 32-Page CoT Loops and English-Dominant Reasoning(5 posts)→

Original post →

More from Models

Models channel →