4 Comments
User's avatar
Nibmeister's avatar

I use the $20/month plan for Claude and I get put on timeout occasionally despite using Sonnet 5 for almost everything. Instead of paying for a higher priced plan, I got a $20 plan for ChatGPT and the $30 Grok plan, as I use these two to check Claude's work and as a red team. For me, $70/month works out better than a $100/mo Claude subscription.

Shimshon's avatar

That also might explain the two weekly token resets this week, on Tuesday, and again today.

They were regular occurrences in May and June, tapered off to nothing in July and August, then suddenly twice in a few days.

I was not keeping track of actual usage in detail, but I easily hit the 20x weekly limit every week. I can hit it within 48 hours if I don't pace myself. I had not read the fine print. I would really appreciate getting the 20x I thought I was paying for.

Shimshon's avatar

The one metric I did become familiar with was the hourly window. One pegged window, which is up to five hours of activity, is 20% of the 20x plan. It it possible to use your weekly limit in 25 hours. That's pretty crazy.

Let the price wars commence.

Original H's avatar

I have a 5x Team plan and a 20x individual plan. I'll have to start measuring how many tokens I get out of each. I routinely run out of tokens on both plans both on the session limit and on the weekly limit.

This week, I tried subscribing to the z.ai GLM Coding Pro plan in addition. Ran out of weekly credits the first day there. The output of GLM 5.3 is in my estimation roughly equivalent to Fable for coding tasks based on my day of usage, but the plan is more expensive than Claude Max plans in terms of cost per result.