isn’t that always zero?
if that is the actual cost, then I’m almost at the end of the month and still have zeros for all the API’s
No, that’s always 0, that part doesn’t matter. What we meant is that although they say it’s “FREE,” it still appears to count toward the API usage. What does that mean? If you reach the limit of your plan, you’ll start paying extra, which means it’s not truly free. We just want clarification on whether, even though it shows an API cost, it is not actually counted toward the API usage limit.
Hi, I noticed that the azure api + gpt-5 setting is not supported so far. As the azure api + gpt-4.1 combination is already well supported, I kind of expected gpt-5 to work as well. Could you help me with the issue please?
Can we get the non-thinking versions of GPT-5 please? (would be great if Thinking was just a switch in the model pop up instead of separate models)
The api usage column will still show the actual calculation, what you should look is the “Cost to you” column (last column in the right), if that is zero then it has no cost to you = free.
1 Like
Cursor’s pricing strategy is always so vague and unclear. I suspect this is the only way they can dynamically adjust the monthly bonus allowance based on total user usage and their own profit margins. I actually have doubts about whether it’s truly “free.”
According to my dashboard, the web backend shows $36.72 in usage, while the IDE shows $35.8 in usage. The $0.9 difference represents the auto allowance, so did GPT-5 actually consume my quota?
1 Like
As a Pro Plus user, I’ve just lost a third of my usage because I assumed GPT-5 was free, but it’s not. In the KIND section, it says “Included in Pro Plus,” not “FREE.” I’m not sure why it isn’t free for Pro Plus users, but that’s what I’m seeing. This means I’ve already used $24.18 + $1.13 of my API cost using the so called FREE GPT-5.
Same question — and an IMPORTANT one: are all GPT-5 versions free? Like high/low? Or just the regular GPT-5?
I personally noticed that GPT-5 doesn’t create to-do lists or provide useful headers of what it’s doing in each step when I give it a multi-step process to complete.
Feels like a big downgrade from Sonnet 4 right now.
2 Likes
Why can’t I see the GPT-5 model in Cursor? I’ve already updated to the latest version, and I don’t even have the Claude 3.5 Sonnet-3.7 model. I’m on the Pro plan too — can anyone explain this?!
Yes, I checked.
1 Like
Hey, is any API key enabled for you? If so, try disabling it and restarting Cursor. This should refresh the list of models. Anyway, you can try adding these models manually.
Bro, which part of “Cost to you = 0” you did not understand? It has zero cost to you, the chart just show the api cost for comparison. Also, that api cost is what you’ll be paying if you use the api directly, but since you use it in cursor then actual cost is zero. But I get it, cursor should be more clear, that they should not affect the usage limit for you guys with the pricing model. Me under old pricing, it is more clear there, it does not consume my 500 fast request, meaning it is indeed free to use.
I haven’t found any information yet about each of them. Can someone explain to me or share a website that explains why there are so many GPT models? 
Specifically, what each one does — because I’m not even sure which is better, GPT-5 or GPT-5 Hard.
I haven’t enabled any API keys yet. When I tried enabling Anthropic API Key and Google API Key, some older model versions appeared, like claude-3-opus-20240229, claude-3-sonnet-20240229, and gemini-1.5-flash. However, I can’t find where to enable the OpenAI API Key, which seems particularly strange. Currently, I can only use Auto mode, which automatically uses Claude 3.5 Sonnet for me. My Cursor version is up-to-date. Now, if I want to use GPT-5, is there any way to make the model appear in my list?
The OpenAI API key settings are at the very top, but if your model list still doesn’t update, you can add them manually anyway.
hey @deanrie while we are on gpt 5 topic, are there any plans to replace the gpt 4.1 or any under-performing models in auto with gpt 5 mini or gpt5 low or any of the versions ?
i personally dont see any reason for non thinking but i agree better model selection would be nice, right now we have SIX models just for gpt5 in the selection
1 Like
anything ending in -fast is… well faster, not sure how if api provider is different maybe? @condor
-low has low reasoning (weakest)
-high has high reasoning (best)
gpt-5 and gpt-5-fast have medium reasoning
edit: sorry for tag, its all in the hover (would help if other models had it too :p)
1 Like