# Anyone else seeing dramatically lower Work usage with Sol 6.1? — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/9d8b1c751cfcdd42

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-09-30 10:23 ingested 2026-10-01 02:38

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1wu1apw/anyone_else_seeing_dramatically_lower_work_usage/)

## retained text



```
I've been using Sol 6.1 Medium in Codex this morning while vibecoding a Commodore 64 game in 6502 assembly, and I'm curious what other people are seeing with the new model's usage.



Admittedly, a C64 game may not exactly be the hardest workload you can throw at a modern coding model but I've been working for roughly an hour and my weekly Work usage hasn't dropped at all.



During that time Codex has done several real repo tasks: modifying the scrolling/background renderer, rebuilding and testing the PAL binaries, running regression/stress checks, launching and inspecting the game in VICE, and cleaning up some fairly timing-sensitive VIC-II code.



It's also been doing a fairly competent job. Sol 6.1 Medium has handled everything I've asked so far without me needing to bump it to High.



Earlier this week, under the previous model/accounting, I'm fairly sure multiple tasks of this size would have cost me 1-2% of the weekly allowance. Nothing it's done this morning has caused the meter to move at all.



I'm not using 1.5x speed, so that could partly explain why it might be sipping tokens at the moment.



I know there was talk about changing the way credits/usage are calculated so allowances stretch further. I'm wondering whether I'm actually seeing the effect of that, whether these particular tasks are simply cheap for 6.1. Or maybe the usage meter is lagging?
```



document fetched 2026-10-01 00:23 · canonical https://reddit.com/r/ChatGPT/comments/1wu1apw/anyone_else_seeing_dramatically_lower_work_usage
