You can run it yourself, so that rules out itβs just Indian people like the Amazon no checkout store was.
Other than that, yeah, be suspicious, but OpenAI models have way more weird around them than this company.
I suspect that OpenAI and the rest just werenβt doing research into less costs because it makes no financial sense for them. As in itβs not a better model, itβs just easier to run, thus it makes it easier to catch up.
Mostly, Iβm suspicious about how honest the company is being about the cost to train the model, thatβs one thing that is very difficult to verify.
Does it matter though? Itβs not like you will train it yourself, and US companies are also still in the dumping stage.
It does, because the reason the US stock market has lost a billion dollars in value is because this company can supposedly train an AI for cents on the dollar compared to what a US company can do.
It seems to me that understating the cost and complexity of training would cause a lot of problems to the states.
They explained how you can train the model in order to create a similar A.I. on their white paper: https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf