Pi using about a quarter of Codex CLI's tokens on the same Qwen model is what … · @yinpengxuan · TwKit

@yinpengxuan

Pi using about a quarter of Codex CLI's tokens on the same Qwen model is what stuck with me. On a local model tokens are basically wall-clock time, so harness bloat is time you spend waiting. Five small tasks is a tiny sample though, curious if the gap holds on a real repo.

曝光 25 · 评论 1 · 点赞 1 · 书签 0