跳到正文
RCreddit.com·

How do you figure out which OpenAI usage is worth the cost?

时间与来源

时间显示为 UTC

显示时区:UTC

本地时区尚不可用,暂时显示 UTC。

发布当时偏移:UTC+02026年9月8日 15:49 UTC

收录当时偏移:UTC+02026年9月9日 15:00 UTC

发布
2026年9月8日 15:49
收录
2026年9月9日 15:00
来源类型
开发者社区
档位
社区
信源状态
同步延迟

档位是按信源手工设定的编辑判断,不是逐条打分。

讨论趋势

暂无对比
最近 24 小时与此前 24 小时的快照均值对比 · 7 天曲线

百分比基于采集到的讨论信号,不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。

正文

We’ve gotten to the point where just looking at our total OpenAI spend every month doesn’t tell me much anymore since some of the increase is completely fine because we’re simply doing more with it but there are also workflows using larger models, longer context and way more tokens than others then you have retries and agent runs where the cost of one task can vary quite a bit depending on what happens along the way.

Seeing that we spent more this month than last month isn't really enough to tell me whether usage is getting better or just more expensive and I’m starting to think we need to look at this closer to the workflow level and compare what the tokens are actually accomplishing instead of treating all OpenAI usage as one bill.

If any of you guys are running a decent amount through the API. What are you actually tracking? Cost per workflow/task, tokens per successful run, spend by team/project, retries, model usage or something else cause I'm trying to work out what metrics are useful before we build a bunch of reporting that doesn't get looked at.

来源·reddit.com