Discerning which models are premium versus fast.

Discussion
1

Hello,

I’m finding that using Fast models for simpler tasks and Premium models for complex ones helps avoid the recent slow request issues. However, I’d like to use the best Fast model for those simpler tasks.

With many models available, I’m unsure which are Premium and which are Fast. For example, is DeepSeek v3 a Premium model? I’ve confirmed Gemini Flash models are Fast, but I’d like a clearer way to distinguish between them without contacting support.

In other software, Premium models are marked with a premium tag.

Is there a reference for which Cursor models are Premium and which are free?

Thanks!

1 Like
2

Currently there are premium models which are claude 3.5 sonnect and gpt. Currently the deepseek model is just a test version so it can be deleted at any time, but if it is officially released, there is a high chance that it will be in the premium model.

3

Ah I see, so essentially the only models that are Premium are Claude 3.5 Sonnet and GPT 4o? All other models are considered fast and do not count towards the Premium budget? But these fast models are more changeable?

4

yes, but deepseek is not clear whether it is fast or pro

1 Like
5

Deepseek is not currently a premium model as it’s cost per token is much lower than the premium models.

2 Likes
6

@danperks Is there a privacy policy change on deepseek-v3?

7

No, we use Fireworks to run the DeepSeek v3 model, and our privacy agreement remains in effect for it.

1 Like
8

Awesome, thanks for the reply!

1 Like
9

Deepseek R1 has quite low costs as well, when it gets implemented into Cursor, will it also not be a premium model?

10

My wild guess is the team is reevaluating with the board their pricing model. DeepSeek changed the rules completely

1 Like
11

Deepseek is literally another chatgpt moment , cursor is suddenly very profitable given the price of deepseek and quality of it.

1 Like