o3 would also be superb!
1 Like
We’re working to add DeepSeek. Their API platform page has been down for days now.
We’ll get it up as soon as it’s available!
2 Likes
@admin_mike @nathaniel Looks like Qwen 2.5 by Alibaba is free (and performs better) 
4 Likes
Guys, the platform seems to be up already 
.
Now that their platform is up, we’ve set the wheels in motion!
3 Likes
Make sure you don’t end up in jail…we need Pickaxe more than DeepSeek 
2 Likes
Maybe Google Gemini 2.0 (price per million tokens is $0.45) is a good alternative. Recent benchmarks showed it performs slightly better than GPT 4o.
2 Likes
Closing out this feature request thread as it has now been implemented.
We have removed DeepSeek, as it has come to our attention that they are training on data collected through the API. Existing users will be able to continue using it, but once they switch away from it, they won’t be able to select it again.
4 Likes
That’s unfortunate… I like DeepSeek. Can you help me understand what the problem is with DeepSeek Using the data collected through the API to train their models? The fact you took the option away seems intense to me. Could it be an option with a warning that allows pickaxe users to make that judgment call whether or not they want to use DeepSeek knowing they are using our data to train? Or is there something I’m missing here?
Deepseek is the best model for translation to the Chinese language, no other models come close. Can we have it back but probably with caveats? Because some customers really do not care whether they train on it or not.
1 Like
@nathaniel, @james.yeoh has a point. In Asia/SEA the intersection of Chinese and English-speaking businesses is very common. It makes sense to offer users an LLM that excels in Mandarin.
Can I suggest Pickaxe looks into the distilled version of DeepSeek from Groq? I understand its decentralized from High-Flyer so data concerns that come with the original DeepSeek should not be an issue with Groq.
Here is another potential workaround (DIY via OpenRouter):
1 Like
Thanks Ned. Allowing us to choose from OpenRouter will be very good!
1 Like
That’s possible via a Make action and scenario.
The automation scenario can look something like this:
Make webhook >>> OpenRouter >>> Webhook Response
All you have to do is sign up to OpenRouter, add credits, and choose your LLM of choice, and finally, connect the OR module in Make/n8n to your account.
1 Like
Thanks Ned, that means I can do this with n8n too. I’ll give it a try 
I currently use open router to access DeepSeek to run all my automations on pickaxe. But in your post I want to be clear, are you showing a workaround for using DeepSeek with pickaxe or just how to use DeepSeek through an API that doesn’t train on your data without pickaxe?
@enochleffingwell , how do you use open router on Pickaxe? Are you using Ned’s method?
Yeah I’ve been using n8n to access openrouter. I tried switching my nodes to Pickaxe Community Nodes in n8n but they have been a buggy experience for me and causing several disruptions. I’m about to stop using Pickaxe in n8n entirely. I’ll go back to using the custom HTTP requests first and hope that fixes my problems…