We’re not far into 2025 and already there’s been some big news in the AI world – mainly that China’s DeepSeek AI has caused a bit of a stir by claiming that its LLM has similar levels of performance to ChatGPT – but has been created and run on a much lower budget, both in terms of money, resources and technology.
According to this BBC news item DeepSeek’s R1 base model V3 reportedly cost around $6m – whereas OpenAI’s GPT-4 cost around $100m.
Also DeepSeek claims to have setup their LLM on lower specced NVIDIA chips than the higher powered – and most expensive – ones. On the day the story came out about DeepSeek, NVIDIA’s stock lost around 17% of its market cap, a massive amount of money, around $600 billion!