On August 3, the overseas developer open-source project team OpenCode disclosed a remarkable spike in the utilization of DeepSeek V4 Flash on its platform, with the daily processing volume soaring to 8 trillion tokens. Out of this total, 5 trillion tokens were utilized within the free trial limit, while developers opted for the paid service for the remaining 3 trillion tokens. Concurrently, data from the model API distribution platform OpenRouter indicated that DeepSeek V4 Flash emerged as the leader in the previous week (spanning from July 27 to August 2), registering a weekly usage volume of 7.22 trillion tokens. In light of DeepSeek's impressive showing, OpenAI responded by slashing the price of GPT-5.6 Luna by 80% on July 31. However, this move failed to impress developers, with some explicitly expressing their preference for DeepSeek's pricing strategy. Even after OpenAI's price reduction, the cost per task for DeepSeek V4 Flash via its proprietary API remained roughly 60% lower than that of OpenAI. This was largely attributed to DeepSeek's substantial cache hit discounts, which amounted to approximately 98%.
