DeepSeek API 定價
適合技術團隊DeepSeek 依使用量計價，而非依月計價；每個選項的費率列於下方表格中。 已於 2026年9月29日 對照官方定價頁面查核。
Very low-cost frontier-class models. Since 10 Sep 2026 the workhorse is V4.1 Flash on the deepseek-flash API id at $0.15–$0.30 in / $0.60–$1.20 out per 1M (off-peak/peak), with image input and MIT open weights; V4 Pro requests route to V4.1 Flash from 14 Sep 2026.
DeepSeek is still one of the lowest-cost ways to run a capable model through an API, with both OpenAI- and Anthropic-compatible endpoints. Pricing now depends on the time of day, so batch work you can schedule off-peak costs half as much.
方案與價格
| 方案 | 價格 | 你獲得什麼 |
|---|---|---|
| deepseek-flash我們的選擇 | 依用量計費DeepSeek-V4.1-Flash. Input cache-miss $0.15/1M off-peak, $0.30 peak; input cache-hit $0.003/$0.006; output $0.60/$1.20. Peak 01:00-04:00 & 06:00-10:00 UTC Mon-Fri; off-peak is 50% of peak. Legacy ids deepseek-v4-flash and deepseek-v4-flash-vision-exp still accepted, route here. | 1M context, 384K max output, vision supported, concurrency limit 2500
|
| deepseek-v4-pro | 依用量計費NOT retired. Rates (DeepSeek-V4-Pro-0813): input cache-miss $0.66/1M off-peak, $1.32 peak; cache-hit $0.022/$0.044; output $1.98/$3.96. Peak 01:00-04:00 & 06:00-10:00 UTC Mon-Fri. | 1M context, concurrency limit 500, no vision
|
選擇此等級：deepseek-flash
The Flash model is the cheapest DeepSeek option and covers most workloads. Schedule batch jobs off-peak to halve the bill.
表現良好之處
- Very low per-token prices compared with the frontier labs
- Off-peak hours bill at half the peak rate
- Cache hits cost a small fraction of fresh input
- Drop-in compatible with OpenAI and Anthropic API clients
需要留意之處
- Peak and off-peak billing makes cost forecasting harder
- Model names and routing change quickly; check which model an ID now serves
- Data-location requirements rule it out for some businesses
以下情況可略過
- Your data must stay with a provider in your own region
- You need a chat subscription rather than an API
- You cannot tolerate model IDs being retired and rerouted at short notice
最適合
DeepSeek 定價相關問題
- DeepSeek 在 2026 年要多少錢？
- DeepSeek 依使用量計價，而非依月計價；每個選項的費率列於下方表格中。 已於 2026年9月29日 對照官方定價頁面查核。
- Does DeepSeek API pricing change by time of day?
- Yes. DeepSeek bills a peak rate and an off-peak rate, and off-peak is half of peak. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays; every other hour bills off-peak.
- What happened to deepseek-v4-flash?
- DeepSeek retired the deepseek-v4-flash model ID on 2026-09-10. Calls to it now route to V4.1 Flash and bill at the V4.1 Flash rates.
- Is DeepSeek's API compatible with OpenAI and Anthropic clients?
- Yes. DeepSeek offers OpenAI-compatible and Anthropic-compatible endpoints, so most existing client libraries work by changing the base URL and key.
- How do I pay less for the DeepSeek API?
- Run non-urgent jobs in off-peak hours, which bill at half the peak rate, and reuse prompt prefixes so input is billed at the much lower cache-hit rate.
Dapols team 於 2026年9月11日 審查。 回報價格變動
DeepSeek 適合你的企業嗎？
回答幾個問題，看看 DeepSeek 是否符合你的工具組合與預算。