API、托管或本地运行
DeepSeek-V4-Pro
DeepSeek API (OpenAI-compatible)、DeepSeek API (Anthropic-compatible)、Responses API
接入说明：DeepSeek API V4 Flash 0731 update ↗选择要核对的信息，直接进入对应章节；各项事实均保留版本、条件和来源。
7/14 项有公开值 · 当前版本 DeepSeek-V4-Pro · 数据截止 2026-10-20
页面只展示已审核 Claim；待审核数据不会被当作公开事实。只展示已审核事实；未公开、冲突和过期状态不会被补猜。
只判断选型资料是否足够，不评价模型能力高低。
所有必填事实均通过状态、来源、证据与时效校验。最早核验时间：2026-09-17。
release_datelifecycle_statuscapability_tagsdeployment_optionscontext_window_tokensapi_idinput_price_usd_per_mtokoutput_price_usd_per_mtokavailabilityDeepSeek V4 Pro 是 DeepSeek 的 DeepSeek V4 模型，年鉴当前记录 1 个版本，最新公开记录为 DeepSeek-V4-Pro。
权重为开放权重，代码为暂无已审核记录，训练数据为暂无已审核记录，部署方式为暂无已审核记录；这些维度分别判断，不能合并成单一“开源/闭源”结论。
当前公开快照的上下文窗口为 1M tokens，架构为 未公开；这里只使用通过审核且来源允许公开采集的 Claim。
6 个维度暂无可公开投影的审核记录
补充来源或线索会进入投稿队列，经编辑核验后才会成为公开事实。当前来源记录 5 个。
技术事实、业务价值、企业治理和证据质量分区呈现
厂商、家族、API ID、日期与生命周期
Dense/MoE、参数、层数、专家与 Tokenizer
预训练、SFT、偏好优化与安全对齐
教师/学生、基础模型、轨迹和限制
权重、代码、数据、许可证、部署与 API
输入输出、缓存、Batch 与部署成本边界
数据、权限、审计、合规、安全与人工兜底
15 条已审核事实 · 全部来自公开来源闸门通过的记录
尚无已审核公开记录的维度：02 能力与任务、07 部署与接入、09 性能与评测、11 业务场景、12 落地与提效。查看资料缺口 ↑
关系只来自年鉴数据库中已登记的归属字段。
下载、API、客户端和更新内容只展示已登记的来源事实；未登记不作推断。
DeepSeek API (OpenAI-compatible)、DeepSeek API (Anthropic-compatible)、Responses API
接入说明：DeepSeek API V4 Flash 0731 update ↗1 个已登记地区 · 1 条版本级记录
The DeepSeek API quick start lists model names deepseek-flash and deepseek-v4-pro, with base_url https://api.deepseek.com for the OpenAI format, https://api.deepseek.com/anthropic for the Anthropic format, and notes that the DeepSeek API is compatible with OpenAI and Anthropic SDKs.
保留发布日期精度、生命周期节点和原始来源，未公开日期不会被推断。
模型版本 · active
DeepSeek 官方公告记录 V4 Preview 于 2026-04-24 上线并开源，明确包含 V4 Pro 与 V4 Flash，API 当日可用。
来源：DeepSeek-V4 Preview Release ↗显示最新核验的 8 条 · 共 15 条已审核事实
The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
1 个版本 · 每项可回到来源与技术解释
高置信度，已绑定来源与证据。
条件：Model name to pass in the request body; served by model version DeepSeek-V4-Pro-0813. The API is OpenAI-compatible at https://api.deepseek.com and Anthropic-compatible at https://api.deepseek.com/anthropic.
来源：DeepSeek API Models & Pricing (English) ↗The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
高置信度，已绑定来源与证据。
条件：Peak-hour cache-hit input price, USD 0.044 per 1M tokens; off-peak is USD 0.022. The discount applies only to tokens that hit the context cache.
来源：DeepSeek API Models & Pricing (English) ↗The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
引用文本与接口返回都只包含已审核公开事实
DeepSeek DeepSeek V4 Pro｜最新版本 DeepSeek-V4-Pro｜已审核事实 15 条｜公开来源 5 个｜数据截止 2026-10-20｜AI 年鉴｜https://ainianjian.com/models/model.deepseek.deepseek-v4-pro
公开接口返回该模型的版本、已审核事实 15 条及其来源信息，可直接用于脚本或对比。
https://api.ainianjian.com/api/models/model.deepseek.deepseek-v4-pro ↗curl -s 'https://api.ainianjian.com/api/models/model.deepseek.deepseek-v4-pro'
技术规格使用当前版本已审核且来源允许公开采集的事实；模型级开放信息与跨版本地区记录分别标注范围。请按每项适用范围核对，未登记字段显示「未公开」，不使用估算值填充。
The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
The official release note is titled DeepSeek-V4-Pro GA Release, dated 2026/08/13, and states that DeepSeek-V4-Pro is launching that day with agent upgrades, flexible reasoning effort and native Responses API support. It adds that V4 Pro is available via API with unchanged model names, and that the new peak and off-peak pricing takes effect at 16:00 UTC on Aug 16, 2026.
The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
The DeepSeek API quick start lists model names deepseek-flash and deepseek-v4-pro, with base_url https://api.deepseek.com for the OpenAI format, https://api.deepseek.com/anthropic for the Anthropic format, and notes that the DeepSeek API is compatible with OpenAI and Anthropic SDKs.
DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
条件：Explicit public claim projection from agent; this is a documented capability tag, not a benchmark or quality conclusion.
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
条件：Hosted options only. DeepSeek publishes no downloadable weights for this version, so self-hosting is not recorded.
来源：DeepSeek API V4 Flash 0731 update ↗The DeepSeek API quick start lists model names deepseek-flash and deepseek-v4-pro, with base_url https://api.deepseek.com for the OpenAI format, https://api.deepseek.com/anthropic for the Anthropic format, and notes that the DeepSeek API is compatible with OpenAI and Anthropic SDKs.
高置信度，已绑定来源与证据。
条件：Peak-hour list price for cache-miss input, USD 1.32 per 1M tokens; off-peak is USD 0.66. Peak hours are 01:00-04:00 and 06:00-10:00 UTC on weekdays excluding Chinese public holidays.
来源：DeepSeek API Models & Pricing (English) ↗The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
条件：DeepSeek published a GA release note on 2026/08/13 and still lists the model as a current API model; the company publishes no formal lifecycle or deprecation table, so this status reflects GA plus current listing, not a vendor API-status field.
来源：DeepSeek-V4-Pro GA Release ↗The official release note is titled DeepSeek-V4-Pro GA Release, dated 2026/08/13, and states that DeepSeek-V4-Pro is launching that day with agent upgrades, flexible reasoning effort and native Responses API support. It adds that V4 Pro is available via API with unchanged model names, and that the new peak and off-peak pricing takes effect at 16:00 UTC on Aug 16, 2026.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
条件：Peak-hour list price, USD 3.96 per 1M output tokens; off-peak is USD 1.98. Thinking tokens are billed as output tokens.
来源：DeepSeek API Models & Pricing (English) ↗The English DeepSeek API pricing page lists deepseek-v4-pro (model version DeepSeek-V4-Pro-0813) with a 1M context length and maximum 384K output. Per 1M tokens in USD: cache-miss input 1.32 at peak and 0.66 off-peak, output 3.96 at peak and 1.98 off-peak, cache-hit input 0.044 at peak and 0.022 off-peak. Off-peak rates are half of the peak rates; peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday through Friday excluding Chinese public holidays, and all other hours are off-peak.
高置信度，已绑定来源与证据。
条件：The official page says the effective date is subject to a future announcement. Do not apply this multiplier to current prices.
来源：DeepSeek API 模型与价格 ↗DeepSeek states that the API will soon adopt a peak/off-peak pricing policy, with peak hours charged at 2x regular prices for all billing items; the effective date is subject to an official announcement. Peak hours are listed as 09:00-12:00 and 14:00-18:00 Beijing Time (UTC+8).
高置信度，已绑定来源与证据。
条件：Official V4 Preview announcement says V4 Pro is live, open-sourced and available through the updated API that day.
来源：DeepSeek-V4 Preview Release ↗DeepSeek labels the V4 Preview release 2026/04/24, identifies DeepSeek-V4-Pro and DeepSeek-V4-Flash, states that the preview is live and open-sourced, and says the API is updated and available that day.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.
高置信度，已绑定来源与证据。
来源：DeepSeek API 模型与价格 ↗DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro: 1M context, maximum 384K output, thinking/non-thinking modes, JSON output, tool calls, API support, token prices, and concurrency limits.