I measured $20 coding plans: ~33 tasks per 5h on Sonnet, only ~6 on Opus
Sufficient-Storage87 · reddit · 2026-10-25
A Redditor ran the same real 8-file WeasyPrint feature on each $20 plan and read each tool's usage meter (2–4 runs per setup):
- Claude Pro + Sonnet 5.5: 3% per task → 33 tasks per 5-hour window
- Claude Pro + Haiku 5.5: 5% → 20 tasks
- ChatGPT Plus + Codex Sol 6.1: 5.5% → 18 tasks
- Claude Pro + Opus 5.5: 15% → only 6 tasks
Weekly limits barely moved: 1% per task for Opus, 0.5% for Sonnet and Codex. The surprise: Haiku, the 'cheap' model, fit fewer tasks than Sonnet—likely because it's slower and re-reads code more, eating more of the window. Caveats: small sample, one task type, integer-only meters. Bottom line: for feature work on Pro, Sonnet delivers the most per window.
More from coding & agent
- print3d: open-source agent skill turns photos or text into Bambu Lab 3D prints — udmrzn · 2026-10-25
- Dev launches gaming-focused remote MCP server for squad-finding, seeking testers — Rashad8705 · 2026-10-25
- Redditor compares the agent boom: Claude Code, OpenClaw, Hermes vs hosted rivals — Odd-Aside456 · 2026-10-25
- Jev ecosystem explodes: 5 open-source apps from a 7.1s flight-search agent to SQL-in-English — udmrzn · 2026-10-25
- PhD student says Claude now does most of his thesis work, down to the draft — GarrisonLovely · 2026-10-25
- Google's Jon Barron shares his full AI paper-writing workflow with Gemini — lmoroney · 2026-10-25