DeepSeek V4 Flash launches, cutting costs by 60% vs GPT-5.6 Luna

DeepSeek V4 Flash's official API launched on July 31, with AI task costs about 60% lower than GPT-5.6 Luna. OpenRouter's weekly stats show it topped global usage with 7.22 trillion tokens, sparking the 'DeepSeek kill line' discussion.

Users say DeepSeek's extreme cost-effectiveness sets a harsh industry benchmark, pushing Silicon Valley giants to cut prices and threatening mid-tier models. Honor's CMO Guan Haitao praised it as 'Ferrari performance at scooter prices.'

To compete with new Chinese models like Kimi K3 and DeepSeek V4 Flash, major AI firms are lowering prices and boosting entry-level model capabilities.