While the price tag of DeepSeek’s R1 model is attention-grabbing, some experts are urging caution. DeepSeek’s $5.6 million figure for training V3 raises eyebrows because it’s so far below the billions typically required to train competitive large language models (LLMs) like GPT-4. Experts like Daniel Newman, CEO of tech insight firm The Futurum Group, have called the claims “a massive breakthrough,” but they’ve also expressed skepticism over how accurate these cost estimates really are.