Honor CMO Praises DeepSeek-V4-Flash's Disruptive Cost-Performance Ratio
Honor's Global CMO Guan Haitao praised the newly released DeepSeek-V4-Flash API, comparing its high performance and low cost to a "Ferrari at the price of an electric scooter." The model has topped the OpenRouter platform with over 7.22 trillion weekly tokens, sparking discussions about a "DeepSeek execution line" that forces global competitors to lower prices. The extreme cost-efficiency of DeepSeek-V4-Flash is reshaping the AI market, putting immense pressure on mid-tier models and forcing major Silicon Valley tech giants to slash their API pricing. This shift accelerates the democratization of high-performance AI while intensifying global competition, particularly between US and Chinese AI developers. Benchmark reports indicate that DeepSeek-V4-Flash's single-task AI cost is approximately 60% lower than GPT-5.6 Luna. Technically, DeepSeek-V4-Flash is a 284-billion parameter Mixture-of-Experts (MoE) model with 13 billion active parameters and a 1-million-token context window.
## BACKGROUND
In large language models (LLMs), a "token" is the basic unit of text (such as a word or character) processed by the AI, and API pricing is typically calculated per million tokens. OpenRouter is a popular unified API platform that aggregates access to hundreds of AI models from various providers, serving as a key benchmark for developer adoption and model popularity.