Ask HN: Where can I use DeepSeek from an alternative provider?

6 points by hgo 2 days ago

Hi all,

I'd like to use DeepSeek-R1, but the official provider at https://chat.deepseek.com are having issues to accept new accounts. Also, for my own reasons I'm interested in alternatives.

Ideally, I'd like to self-host a chat-interface similar to the UI that both OpenAI and DeepSeek provide, but I'd expect a third-party solution to quite likely be more tailored to my needs. (I'm actually not very impressed with the official UI:s that I've seen.)

I realize that this is probably a quite basic question to someone in the know, but I haven't found a good answer.

Thank you

kingkongjaffa 2 days ago

https://ollama.com/library/deepseek-r1

vunderba 2 days ago

As has been pointed out by other HN users, these are models distilled from DeepSeek-R1 and are based on Llama and Qwen. There's almost zero chance OP has a computer capable of running the full 671b R1 model locally.
Realistically they might be able to run a 4-bit quant of the 70b model though.
2024user 2 days ago

Is Ollama related to the llama models?

szszrk 2 days ago

Also Kagi could be a starting point: https://help.kagi.com/kagi/ai/assistant.html#llms-available-...

They have R1 and Chat V3 from Deepseek.

creamyhorror 13 hours ago

DeepInfra and others listed at OpenRouter.

maxbaines 2 days ago

Fireworks AI is a great option - https://fireworks.ai/models/fireworks/deepseek-r1

adrianmoses 2 days ago

Open WebUI is pretty is to set up and works with Ollama https://github.com/open-webui/open-webui

high_byte 2 days ago

also perplexity.ai (pro subscription)

pestaa 2 days ago

Not rolled out on the phone app, but having the option on desktop so soon was a pleasant surprise.