Huge Discrepancy in "Auto" Mode Limits (Nov vs. Dec). Was policy changed?

Hi everyone,

I have been a Pro plan ($20/mo) user for about a year and STRICTLY and use “Auto” mode for all my requests. I am looking for clarification on how the “fair use” or “usage limit” is being enforced, as I am seeing a massive difference between November and December behavior despite similar usage patterns.

The Issue: In December, I hit my usage limit and was switched to “On-Demand” (paid) pricing after using about $50 worth of resources. However, looking at my logs for November, I used significantly more resources without ever hitting a limit.

My Usage Data (from usage-events logs):

  • November 2025:

    • Total Usage Cost: ~$147.00

    • Total Requests: ~690

    • Avg Tokens per Request: ~450k

    • Result: 100% “Included” (Never hit a limit, never asked to pay).

  • December 2025 (Current):

    • Total Usage Cost: ~$50.00

    • Total Requests: ~140

    • Avg Tokens per Request: ~290k

    • Result: Limit Hit. Switched to “On-Demand” pricing.

My Questions:

  1. Has the policy for “Auto” mode changed recently? It seems I was allowed to use ~7x my plan cost in November (“Included”), but in December, the limit was enforced much strictly.

  2. Is “Auto” mode no longer treated as “Unlimited” (with a slow pool fallback)? It seems that once the dollar limit is hit in Auto mode, I am forced to pay for overage rather than falling back to a slow queue. (you can see the screenshot in the image i attached in the end. its the same with all modes)

  3. Was I simply “lucky” in November, or is the new enforcement on Auto mode intended to be this strict (capping at ~$20-$50)?

WHAT MUST I DO TO CONTINUE USING CURSOR ?

3 Likes

Yes.. they did change their policy in November… auto isnt free anymore, you need to figure out what models you can afford.

3 Likes

The auto mode, was the best on cursor, no auto mode, no fun…
Moving….

2 Likes

It was never practical long term to give unlimited high quality requests for $20/mo. Where are you moving that offers that?

1 Like

Happended today ,
Whats the solution
I dont have a big budget

A reminder: Models | Cursor Docs . In short, you need to decide what models you can afford to use. Personally, if you have a limited budget, use GPT-5-mini, its fast ,a logical model, and goes at $2 per Million tokens. Thats 10 M tokens to work with before more usage is required.. Codex is still buggy IMHO within cursor. And nano is not smart enough for development tasks…

Auto was SUPPOSED to be the budget-friendly option. It’s what has been, and always was, recommended to save tokens. To save costs. Every “plan upgrade” has taken it above and above, but always left auto alone. Able to survive.

Not anymore. That’s pretty ■■■■■■■

Bruh be happy that you got over 100$ last month with just the basic plan, i have the ultra plan and its free usage is capped at ~500$

Auto mode sucks $ as sonnet 4. itsa huge degradation of cursor itself. seems its time to move forward to claude and others.

2 Likes

Sadly, Cursor has done it again. They changed their pricing without proper notice to customers—no email, no newsletter, nothing. Cursor does the job, but this kind of approach is frustrating for customers. Acting this way will only push people away as soon as a new alternative becomes available.

1 Like

Yes, they are known to fraud. I had opted out for their new pricing system and they changed it without notification two months ago. Declined to address the issue and only avoided the subject.

Cursor just hid that post so people can’t be awakened!

1 Like