DeepSeek V4 Flash Vision 是 DeepSeek 推出的对话大模型,发布于 2026-07。DeepSeek V4 Flash Vision is a Chat Model released by DeepSeek, launched on 2026-07.
DeepSeek V4 系列首款原生视觉模型(实验版,8/21 发布、8 月底开源):图片描述、截图 OCR 与图表分析,Terminal Bench 2.1 得 83.9;图片按尺寸换算 tokens 与文本合并计费。峰谷计价同 Flash:表中为高峰价(输入$0.42/输出$1.25 每百万),闲时半价,缓存命中 $0.015。DeepSeek V4's first native vision model (experimental, released 8/21, open-sourced late Aug): image captioning, screenshot OCR and chart analysis, Terminal Bench 2.1 83.9; images tokenized by size and billed with text. Peak/off-peak same as Flash: listed peak $0.42/$1.25 per M, off-peak half, cache hit $0.015.
DeepSeek V4 Flash Vision 目前以付费 API / 平台调用为主,可关注官方是否提供免费试用额度。DeepSeek V4 Flash Vision is mainly available via paid API/platform calls; watch for official free trial quotas.
DeepSeek V4 Flash Vision 的上下文窗口有多大?How large is DeepSeek V4 Flash Vision's context window?
DeepSeek V4 Flash Vision 的上下文窗口为 1M,适合处理长文档与多轮复杂对话。DeepSeek V4 Flash Vision's context window is 1M, suitable for long documents and multi-turn complex dialogue.
开源模型,可通过官方开源仓库或自部署方式使用。Open-source model, available via official repositories or self-hosting.
DeepSeek V4 Flash Vision 适合用来做什么?What is DeepSeek V4 Flash Vision good for?
DeepSeek V4 Flash Vision 的主要能力包括 chat、vision、function、cache,可用于相关场景的创作、分析与自动化任务。DeepSeek V4 Flash Vision's main capabilities include chat、vision、function、cache, usable for creation, analysis and automation in related scenarios.