Documentation inconsistency: Auto mode limit handling needs clarification

Feedback
1

There is a confusing info about Auto in documentation. the Cursor AI assistant confirms it:

Neither https://cursor.com/pricing, not https://docs.cursor.com/account/pricing, and nor https://docs.cursor.com/models#auto doesn't specify whether Auto enabled consumes any tokens/limits. How is it?

Searched Auto enabled tokens limits consumption pricing
Searched Auto model selection token consumption limits pricing

Based on the search results from Cursor’s documentation, I can provide you with information about Auto mode and token consumption:

Auto Mode and Token Consumption

According to the documentation, Auto mode does consume tokens/limits. Here’s what the documentation states:

How Auto Mode Works

Auto mode “configures Cursor to select the premium model best fit for the immediate task and with the highest reliability based on current demand.” It automatically switches between different premium models to optimize performance and reliability.

Token Consumption

When you use Auto mode, it does count against your subscription limits because:

  1. Auto selects from premium models: The documentation indicates that Auto chooses from a pool of premium models (excluding o3), which are all subject to usage limits.
  2. Follows standard pricing structure: Auto mode follows the same pricing structure as manually selecting models. Each request made through Auto mode deducts from your monthly usage budget based on the selected model’s API pricing.
  3. Usage tracking: You can view all requests and token breakdowns on your dashboard, which includes requests made through Auto mode.

What Happens When Limits Are Reached

When you reach your monthly usage limits while using Auto mode, you’ll see the same options as with any other model:

  1. Switch to Auto (which you’re already using)
  2. Enable usage-based pricing to continue
  3. Upgrade to a higher subscription tier

Key Points

  • Auto mode is not a separate free tier - it’s a smart routing system for premium models
  • It consumes tokens/limits just like manually selecting claude-4-sonnet, gemini-2.5-pro, or gpt-4.1
  • The main benefit is convenience and reliability, not cost savings
  • All usage is tracked and counted toward your plan’s monthly limits

So to directly answer your question: Yes, Auto mode enabled does consume tokens and counts against your subscription limits, just like any other premium model selection in Cursor.

OK, but I don't understand why should "When you reach your monthly usage limits while using Auto mode, you'll see the same options as with any other model: Switch to Auto (which you're already using)" help than?

You’re absolutely right to be confused by that! That’s actually an inconsistency or error in the documentation.

Looking at the pricing documentation more carefully, it states:

"Once monthly usage limits are exhausted, you will be notified explicitly and shown a message with three options:

  1. Switch to Auto
  2. Enable usage-based pricing to continue with the same model (charged at API prices)
  3. Upgrade to a higher subscription tier"

But as you correctly point out, if you’re already using Auto mode and hit your limits, being offered “Switch to Auto” as an option makes no sense - you’re already using it!

This appears to be either:

  1. A documentation error - The options should probably be different when you’re already using Auto mode
  2. A system limitation - The same generic message is shown regardless of which model you were using when you hit limits

The logical options when hitting limits while using Auto mode should probably be:

  1. Switch to Auto (already using it)
  2. Enable usage-based pricing to continue
  3. Upgrade to a higher subscription tier

Or perhaps Auto mode might switch you to a lower-tier model or free tier when limits are reached, but the documentation doesn’t clearly explain this behavior.

This is definitely a confusing aspect of the documentation that could use clarification from Cursor. The “Switch to Auto” option only makes sense if you were using a specific premium model (like claude-4-sonnet) and hit limits, not when you’re already using Auto mode.

Switching to auto would make sense if there are any free models, I guess?

That’s a very logical assumption, but based on the documentation I found, Auto mode only selects from premium models - it doesn’t include any free models.

The documentation specifically states that Auto mode:

“keeps you in flow by picking a reliable model from the pool above (excluding o3)”

And when referring to “the pool above,” it’s talking about premium models like:

  • claude-4-sonnet
  • claude-4-opus
  • gemini-2.5-pro
  • gpt-4.1

The documentation also mentions that Auto mode “configures Cursor to select the premium model best fit for the immediate task” (emphasis mine).

So unfortunately, switching to Auto when you hit limits wouldn’t help because:

  1. Auto only chooses from premium models that consume your usage budget
  2. If you’ve hit your limits, Auto mode would still be subject to those same limits

This makes the “Switch to Auto” option when already using Auto mode even more nonsensical. It seems like either:

  • The documentation/UI has an error
  • There’s some fallback behavior in Auto mode that isn’t clearly documented (like falling back to a basic model)
  • The system shows the same generic options regardless of context

Your reasoning about free models making sense for the “Switch to Auto” option is sound - that would be the logical behavior. But based on the current documentation, Auto mode appears to be exclusively for premium model selection, not a fallback to free tiers.

This definitely seems like something that needs clarification from Cursor’s team!

3

I’ve just found the 1st info about Auto being free in Dashboard/Usage:

Pro comes with at least $20 of included usage per month. We work closely with the model providers to make this monthly allotment as high as possible.

Auto usage does not count towards your usage limits.

This should be mentioned also in the pricing documentation.

1 Like
4

I think developers should try using LLM in their work. For example, they could install a Cursor IDE for themselves and ask the Agent inside to analyze the consistency of their documentation. For only $20/month!

2 Likes
5

The more I am diving into the pricing thingy, the more it seems as a joke tbh.

  1. Nor me neither the Cursor AI assistent were able to find API pricing rates for individual models in the documentation
  2. In https://cursor.com/pricing Individual plans, I see only Hobby, Pro, and Ultra. However, https://docs.cursor.com/account/pricing, AI assistent, and many forum topics also speaks about Pro+. But I don’t have any option to switch to it :sweat_smile: Actually, I cannot switch neither to Ultra (not that I want to, just saying) as Billing & Invoices/Update subscription only shows me the Current subscription Cursor Pro.

Is it just me? Maybe some geo restrictions as I am Czech = EU. Or maybe some changes underway.
Or the Cursor team collapsing…

UPDATE: I’ve realised it is possible to switch to Ultra via Get Ultra at https://cursor.com/pricing

6

The Pro+ missing from https://cursor.com/pricing have been discussed in topics for at least 7 days. Which is - given the recent turmoil around the pricing change - really shame!

8

Hey, let me clarify for you:

  • Auto is unlimited on every plan, and has no effect on your usage allowances each month.
  • Auto does route between frontier models, like the ones you have listed, but this still doesn’t consume any usage as we intelligently route your request based on the capacity of the models, which makes it more affordable for us to serve
  • I don’t believe we suggest switching to Auto if you are already using it, as you should never see any usage errors in the first place - this is likely an AI getting this wrong in your post
  • We do not list the exact costs for each LLM directly, but these are charged at ‘API pricing’, which means you can check the exact costs on the website for the model you have chosen.
  • We believe most users may find they have all the usage they need with Pro, which is why Pro+ isn’t offered on the website. You would be offered it if you hit any usage limits in the editor.
  • We don’t have a button to upgrade within the cursor.com account page, but I think this is a good idea so I’ll pass this to the team!
2 Likes
9

I think a few small clarifications would mostly resolve the negative impression towards both auto mode and the new pricing plan:

  1. Clarify which models are included in the auto pool, updating as this changes over time (I think this used to be listed in the docs?).
  2. Display which model generated any given response when in auto mode
  3. Confirm whether model performance in auto model is intended to match the performance of the same model when selected directly

Assuming auto mode doesn’t serve quantized or otherwise handicapped versions of the normal models, it seems like the negative impression users have of auto mode is just placebo. When users get a bad response and see the model is Sonnet 4 they fix their prompt or context and try again, but when they get a bad response and see “Auto” they think “Cursor must be secretly serving me Haiku”.

Due to the pricing change, most of Cursor’s value proposition has shifted to auto mode. Unlimited Gemini 2.5 / GPT 4.1 / Sonnet 4 is much better than 500 requests in my opinion! But the perception is more like “Unlimited Haiku”. If users know they can trust auto mode, the perceived value of the Pro plan will go up a lot, and you won’t see nearly as many complaints about the pricing changes.

2 Likes
10

Can you explain please why is the auto charged here? if auto is also charging my usage, and is unlimited can you clarify what exactly does this usage indicate?

image
image1830×722 92.8 KB

3 Likes
11

I changed the topic back to UNSOLVED, this needs to be clarified.

12

Cost to you = $0

13

I understand that I haven’t paid for this, but it’s still unclear to me how the $60 limit is calculated. For example, if both Claude and “auto” usage are shown—does usage in “auto” count toward the limit or not?

It would be much clearer if my current usage was displayed in a way that distinguishes between unlimited and limited usage. Ideally, there should be a dashboard that shows exactly how much of the limited usage I’ve used so far, how much is left for the month, and which features or models are counted toward the monthly limit.

Previously, when usage was measured by requests, the dashboard showed how many requests I had used. I think something similar should be available now for the new usage tracking.

14

If I have Pro+ ($60/m limit)

And I’ve incurred expenses

  • $30 Claude 4
  • $10 Gemini
  • $20 Auto
    = Total display as $60.

Does that mean in truth I actually still have $20 of remaining balance? ($60-$30-$10). Or do I actually have $0 remaining balance?

15

What they’re doing is creating a habit for developers to become dependent on Cursor. People will now upgrade to the $50–$60 plan, and once the data shows that the majority have switched to the Pro+ plan, they’ll introduce new limits and then push users toward the new $100–$120 plan. Eventually, they’ll force users onto the $200 plan.

16

Yes, but as both @Da-Vinci and @mokhir mention, it is not clear whether it actually affects remaining balance.

17

Maybe you will be interested in this topic You've hit your usage limit (using only Auto) - #3 by Ex_Noer

1 Like