How many real coding tasks fit in one 5-hour window on the $20 plans? I measured it: Sonnet ~33, Haiku ~20, Codex ~18, Opus ~6
A user measured the number of coding tasks achievable within a 5-hour window on various $20 AI plans. Claude Pro's Sonnet 5.5 completed approximately 33 tasks, using about 3% of the weekly limit per task. Haiku 5.5 managed around 20 tasks, while ChatGPT Plus's Codex Sol 6.1 completed about 18 tasks. Claude Pro's Opus 5.5 performed the fewest at roughly 6 tasks, consuming about 15% of the weekly limit per task. Sonnet 5.5 offered the most work per window for feature writing.
This report uniquely quantifies the actual coding tasks completed per 5-hour window across different $20 AI plans, unlike other comparisons that focus on feature sets or raw performance metrics.
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10月24日 17:58 UTC
收录当时偏移：UTC+02026年10月24日 21:00 UTC
- 发布
- 2026年10月24日 17:58
- 收录
- 2026年10月24日 21:00
- 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
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。
I wanted a straight answer to "what does $20 actually get me" for coding, so I ran the same task on each setup and read the usage meter before and after.
Setup - Task: a real 8-file feature in WeasyPrint (the same one from my earlier token-saver test) - Every model on its default settings, nothing tuned. Opus runs on medium effort by default in Claude Code, Codex Sol 6.1 on medium in Codex - Claude Pro ($20) for the Claude models, ChatGPT Plus ($20) for Codex - Usage read from each tool's own meter, 2–4 runs per setup
setup per task tasks per 5h window Claude Pro, Sonnet 5.5 ~3% ~33 Claude Pro, Haiku 5.5 ~5% ~20 ChatGPT Plus, Codex Sol 6.1 ~5.5% ~18 Claude Pro, Opus 5.5 ~15% ~6 Weekly limit barely moved per task: about 1% for Opus, about 0.5% for Sonnet and Codex.
The surprise: Haiku is the "cheap" model and still fit fewer tasks than Sonnet. My guess is it took longer and went back over the code more to get the same feature done, so it ate more of the window even at a lower rate.
Caveats - 2–4 runs per setup, so these are estimates, not exact numbers - One task type. A small bug fix or a huge refactor would shift the numbers - The meter only shows whole percentages, so the small ones (3%, 5%) are rough - Defaults only. Opus on low effort costs a lot less; I tested that separately and will post it
If you're on Pro and mostly writing features, Sonnet gets you the most work per window by a wide margin.