Qwen 2.5 coder 32b?

Discussion
1

so qwen just dropped a new 32b model that is better then sonnet-3.5-20241022 in some benchmarks which is pretty insane if true, will cursor be adding support for this?

image
image1811Ă—902 154 KB

6 Likes
2

+1

Should be really quick since it runs well locally, already.

3

Today I spent the whole day. To understand how to connect this model to the cursor. I couldn’t find any information. We are really waiting.

Or maybe someone knows how to connect this model to the cursor? Deploy locally or elsewhere.

It showed itself better than the latest Sonnet model! On a difficult issue. Need to integrate.

4

You can use this model through openrouter.ai. You need to enter your API key, change the base URL as shown in the screenshot, and add the model to the list of models.

image
image1196Ă—852 45.4 KB

image
image1668Ă—518 48.7 KB

3 Likes
5

Would love to have this model integrated in Cursor! Maybe “priced” the same as Haiku 3.5. Not sure what makes sense and what current API costs for Qwen 2.5 coder 32b are

6

they could maybe host it themselves since its a small model, ive been using on my 4090 and it works great

2 Likes
8

Hi, I tried to use this method as you did in these instruction.
I cannot use Claude 3.5 Sonnet now.
I am Cursor Pro user…

9

Hey, what error are you getting? I assume you’ve enabled your API key and are trying to use the Claude model from your subscription. If that’s the case, you need to disable your API key. Whenever you need to use the model from your subscription, you should disable the API key. You can also use the shortcut: cmd + shift + 0 or ctrl + shift + 0.

10

{228AFD0A-3B7A-4FB8-B88A-8EA1D3697367}
{228AFD0A-3B7A-4FB8-B88A-8EA1D3697367}387Ă—242 8.33 KB

It shows like this.

I have to disable my OpenAI API key connect through Openrouter.ai so I can continuting use the models.

1 Like
11

Hey, this is necessary as Cursor does not know what models you want to be sent to OpenRouter, so all your queries are sent there. We are hoping to improve the use of custom LLM providers soon!

1 Like
12

I tried qwen coder with groq, disabled all other models, still it tries to use “default” as a model name. how can I enforce specific model on open ai compatible API

image
image2158Ă—1286 222 KB

13

It would be nice if an API from OpenRouter could be used in Agent Mode.