← sources

ChatGPT Work usage draining unusually fast?

source: reddit (reddit:ChatGPT)published 2026-10-20 12:17ingested 2026-10-21 02:36

retained text

Hey guys, I'm fairly new to agentic coding (about 3 weeks in), and I've noticed my ChatGPT Work usage depleting unusually fast over the last 2 days.

I'm using ChatGPT Work locally on my laptop with GPT-6 Sol on Medium.

Here's what happened:

1. Started at **100% usage available**. Attached my local project and asked the agent to read my [`AGENTS.md`](http://AGENTS.md) file (\~281 lines, 30 KB), inspect 4 existing Jupyter notebooks, and create a new notebook based on them. It generated a notebook with just 8 code cells and one figure (PNG/PDF). **Usage went from 100% to 77% in about 20 minutes.**
2. Asked it to make a minor edit to the notebook, literally around 4 lines of code. **Usage dropped further from 77% to 69%.**

The strange thing is that it wasn't like this last week. I haven't changed my [`AGENTS.md`](http://AGENTS.md) file or significantly changed my workflow.

I understand that reading notebooks, maintaining context, and reasoning across multiple steps can consume tokens, but this seems unusually excessive compared to my experience last week.

**Is anyone else experiencing this recently? Or could something in my setup be causing excessive token consumption?**

Would appreciate any insights, especially from people using ChatGPT Work for research or coding projects.

EDIT: Something is not right with my work repository. Have been using Blender MCP with ChatGPT Work 6.1 Sol Medium; usage is under 5%. 

document fetched 2026-10-21 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1xaql7d/chatgpt_work_usage_draining_unusually_fast