According to Reuters, the U.S.-based research firm Artificial Analysis has reported that the latest iteration of DeepSeek's flagship AI model, V4-Flash, has set a new benchmark for cost-efficiency in global evaluations of well-known AI models. Specifically, it charges a mere $0.14 for processing one million input tokens and $0.28 for generating one million output tokens. On average, the cost per session amounts to just 3 cents, making it over 100 times more economical than Anthropic's Claude Fable 5.