DeepSeek undercuts OpenAI and Anthropic on benchmark running costs
A version of Chinese startup DeepSeek’s flagship AI model is by far the least expensive to run on benchmark tests among well-known models globally and more than 100 times cheaper ...