发布于 2026-08-16 · 赵予安
直接回答
DeepSeek V4 Flash CN 面向中文负载，标价为每百万 token $0.020 / $0.041。上下文窗口和缓存安全仍未公布。 这篇指南面向正在比较模型质量、成本、路由策略和生产上线风险的产品与平台团队。
What shipped
Deepseek V4 Flash CN is listed on the NextModel gateway as deepseek/deepseek-v4-flash-cn. The public slug is deepseek--deepseek-v4-flash-cn. The provider is DeepSeek. The model has streaming capability. There is no first-party CacheSafety row for this exact model in the NextModel bench table, so Safe Hit, Bad Hit, and trap rates are unpublished.
Pricing and context
Input price is $0.020 per 1M tokens. Output price is $0.041 per 1M tokens. The context window is unpublished. The latency band is 1000-3000ms. For comparison, Deepseek V3.2 CN is $0.042 per 1M input tokens and $0.064 per 1M output tokens. Deepseek V4 PRO CN is $0.239 per 1M input tokens and $0.479 per 1M output tokens.
Who it is for
The model is for Chinese-language use cases. Its price sits below both Deepseek V3.2 CN and Deepseek V4 PRO CN. Streaming output makes it usable for incremental response delivery.
What is still unverified
The release date is unpublished in this catalog snapshot. Independent third-party quality scores are not in these notes. The context window is unpublished when the catalog value is 0. CacheSafety rates for this exact model are unavailable.