Suggestion to add deepseek model，thank you very much！
Deepseek models can be run locally, if you have a mind and the hardware to do it.
deepseek api？
1 Like
Need to support api, local hardware can’t handle deepseek with more parameters. Thanks!
“need” is subjective and this has already been discussed. We aren’t going to be adding support for every possible model available. There’s much more going on than merely adding a URL.
Does this mean that you will NOT add an option to use Deepseek via api specifically?
Or does it mean you specifically restrict the use of Deepseek in your product?
Their api conforms to openai api specification, so there is no technical excuse since as far as i understand you do allow to use openai models via api.
The DeepSeek API uses an API format compatible with OpenAI. By modifying the configuration, you can use the OpenAI SDK or softwares compatible with the OpenAI API to access the DeepSeek API.
Just a tiny subset of the OpenAI specification. Anyway, we might add support for OpenRouter but there are currently no plans to directly support the DeepSeek API.
1 Like
I am sorry guys, but the way you ACTIVELY RESIST adding deepseek api support just starts giving me conspiracy theory vibes, like your government telling you not to do it or some financial motivation from openai / anthropic. Which is totally fine, if you just said so.
Also, I have read some other threads and noticed that Jim said that the cost is subjective to each individual. I do not know what he meant by saying this but what I understood was that it was too bad I was too poor to allow myself to use rich people’s model.
The price difference is quite objective, all you have to do is a simple google search to find something like this:
You’re being ridiculous.
7 Likes
Nobody tells us anything and never will.
This price comparison is definitely not useful (like most “simple” Google searches nowadays) as e.g. OpenAI offers lots of models, ranging from $0.1 to $20 per 1 million tokens. Even O3 costs only $2 per 1 million tokens. And DeepSeek doesn’t even support images.
3 Likes
Using the deepseek model in ollama is better supported in DT, but it is a quantized version of the model. In fact, it is also to add a support URL, the workload is not too big, why do not you want to increase it, the user is willing to use on the use, do not want to use also does not affect ah, it does make people some effort.
Maybe you can use javascript to build a connection 
What makes you decide for other people if a workload for them is “not too big”?
Especially as what you ask is not only about adding a URL – customers expect that stuff to be working afterward, even if the URL changes or some other thing happens at Deepseek’s end.
BTW, why all the hoopla with ai in DT? I don’t find it worthwhile at all, so I’m still on the older DT.
JavaScript with the Obj-C bridge (NSURLSession) to call DeepSeek directly
Because there are thousands more users than just you 
PS: There isn’t a lot of hoopla about it. While many people are excited about it – and that’s their right to be – we handle the topic in honest and measured ways. And we also have people who don’t want to use it at all, and they’re correct for themselves as well.
Well now, I think an investigation of the tone of the last few months of this forum would reveal something that is at the very least hoopla-adjacent.
I wonder if one of the chattery AyAye agents could sum up the tone of this forum and compare it to the same period last year. Methinks hoopla would be a word they might use.
1 Like