ThinkingCap · Qwen 3.6 27B

Token-efficient reasoning — same answers, ~50% fewer thinking tokens

connecting…
stream the trace

Chat with ThinkingCap-Qwen3.6-27B

A finetune of Qwen 3.6 27B that keeps answer quality while cutting thinking tokens by ~50% on average (and >90% in best cases). Ask anything — reasoning, math, code, or upload an image.