Estimate DeepSeek token costs.

Compare peak and off-peak costs for your own input, cached input and output totals.

DeepSeek API pricing calculator

Estimate the same token volume at peak and off-peak rates. Enter totals for any period; cached input is part of total input.

The peak and off-peak figures each apply one rate to the entire entered volume. Actual cache hits and timing can differ. Peak is 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays; other times are off-peak. Token-only estimates exclude tools; all estimates exclude taxes and credits. Check current DeepSeek prices (checked 31 October 2026).

See peak windows in your time zone

Browser time zone: unavailable. The selected date is in UTC; local dates can differ.

Chinese public holidays are off-peak and are not detected here. Check the current DeepSeek schedule and holiday dates before scheduling.

A mixed estimate assumes the same peak share for uncached input, cached input and output. Use actual token totals by period if your usage differs. Search and Fetch fees do not change by time of day.

Include reasoning in output tokens; DeepSeek enables thinking by default. Search and Fetch each cost $0.75 per 1,000 billable successful requests. Server-side calls through DeepSeek Harness, Claude Code or Codex start per-request billing on 1 November 2026 at 00:00 Beijing time. Only enter calls billable for your situation. Check tool billing rules.

Rates used in the estimate

Direct DeepSeek API rates in USD per million tokens, checked 31 October 2026. A cached token is part of input, not an extra token.

ModelPeriodCache missCache hitOutput
deepseek-flashPeak$0.3$0.006$1.2
deepseek-flashOff-peak$0.15$0.003$0.6
deepseek-v4-proPeak$1.32$0.044$3.96
deepseek-v4-proOff-peak$0.66$0.022$1.98

How to use the estimate

Use the same time period for each field. If you do not know cache hits yet, enter zero to model the uncached case; this is an assumption, not a prediction. To estimate an actual completed request from its usage JSON and your own rates, use the token usage calculator. Cache behavior and account charges can differ from a planning scenario.

Peak and off-peak results price the whole workload in each period. Optionally enter the peak share to estimate a mixture; the same share is applied to all token types. Search and Fetch are separate per-request charges. Read the Search and Fetch billing notes. For your API key and base URL, see the DeepSeek setup guide.

Official DeepSeek model and tool pricing · DeepSeek cache-hit rules