I need help about the api url please | OpenRouter | Page 1

dusty holly Apr 5, 2025, 6:22 AM

#

and after reopen the browser, it became something like this

mint hedge Apr 5, 2025, 6:30 AM

#

You didn't typed in the full model in the Model tab. It's

deepseek/deepseek-r1:free

While you just typed in

deepseek

dusty holly Apr 5, 2025, 6:57 AM

#

mint hedge You didn't typed in the full model in the Model tab. It's deepseek/deepseek-r1:...

thank you, but right now I was using v3 model, should I still type the same thing though?

#

oh! I got it fixed by copy the name from the model result directly, somehow... the original link stil works. thanks!

#

I was using v3 model so I just copu and paste "deepseek/deepseek-v3-base:free" and it works, turns out the name was important, leasson learned

mint hedge Apr 5, 2025, 7:06 AM

#

Yeah, you can just go to the model page in OpenRouter's website and copy the model name

mint hedge Apr 5, 2025, 7:08 AM

#

dusty holly I was using v3 model so I just copu and paste "deepseek/deepseek-v3-base:free" a...

Also, I don't think you should use the Base version since it's just continues from where you left off, unlike a normal chat/instruct response
Like:

You: Hi there
AI: ! How can I help you today?

#

Either use:

deepseek/deepseek-chat:free
Or
deepseek/deepseek-chat-v3-0324:free

#

👍

dusty holly Apr 5, 2025, 7:10 AM

#

thank you, I tested the base version few times, but I find in the end of the generation it likes to repeat it self, like it would stuck for few seconds and continue to generate last paragraph, I guess that's something you can't get rid off. right?

#

I used the custom prompt recommended by janitorai on reddit, but it wasn't really working well on improving that

mint hedge Apr 5, 2025, 7:13 AM

#

Honestly a lot of factors impact AI responses like temperature, penalties, top-p and prompts

#

So maybe it can be rid off, but I personally haven't used DeepSeek so I don't have any advice

#I need help about the api url please