DeepSeek API pricing is pay-per-token. This post attributes numbers to DeepSeek’s pricing page as of 17 September 2026—not eternal truth. Consumer chat is free with no paid chat tier; that is separate from API. Cover Flash, Pro, peak versus off-peak, and cached input. Always recheck platform.deepseek.com before you budget or invoice.
V4.1 Flash rate card
deepseek-flash (V4.1 Flash), per the 17 September 2026 pricing page: peak $0.30 per million input and $1.20 per million output; off-peak half of that ($0.15 / $0.60). Cached input is much cheaper ($0.006 peak / $0.003 off-peak). Specs cited: 1M context, 384K max output, vision supported, concurrency 2,500. Treat these as a dated snapshot. Peak hours are described as weekdays 01:00–04:00 and 06:00–10:00 UTC—confirm the clock on DeepSeek’s pricing page because it can change. Recheck platform.deepseek.com. Treat every rate and limit as a dated snapshot from DeepSeek or the other vendor docs linked in sources, not as a permanent contract price.
V4 Pro rate card
deepseek-v4-pro on the same 17 September 2026 page: peak $1.32 input / $3.96 output per million; off-peak $0.66 / $1.98. Cached input $0.044 peak / $0.022 off-peak. Same long context story—1M context, 384K max output—but vision is not supported and concurrency is cited lower (500). Pro is the premium text path when you accept higher token rates. Do not mix Pro cache numbers with Flash. Always recheck platform.deepseek.com before contracts. Treat every rate and limit as a dated snapshot from DeepSeek or the other vendor docs linked in sources, not as a permanent contract price.
Peak windows and India jobs
Peak hours are described as weekdays 01:00–04:00 and 06:00–10:00 UTC. India daytime often falls outside that UTC window, so many Indian API jobs land in off-peak—but tell your team to confirm the clock on DeepSeek’s pricing page because it can change. Consumer chat remaining free does not change API math. For Cursor billing (Start India ₹649/mo Cursor models only; Pro $20 Other Models at API rates), see Cursor docs—those are not DeepSeek’s price list. Treat every rate and limit as a dated snapshot from DeepSeek or the other vendor docs linked in sources, not as a permanent contract price.
Related on this site
Frequently asked questions
-
Is DeepSeek API free?
No. As of 17 September 2026, consumer chat is free; the API is pay-per-token. Budget Flash or Pro rates separately from chat. Recheck platform.deepseek.com for the live card. Confirm live details on the vendor docs linked in this brief before you quote a client or ship production traffic. Attribute DeepSeek Flash and Pro figures to the 17 September 2026 pricing page when you reuse them, then recheck platform.deepseek.com.
-
What is off-peak pricing?
For the listed Flash and Pro input/output rates, off-peak is half of peak on the 17 September 2026 page. Cached rates also differ by peak/off-peak. Confirm windows and prices on DeepSeek’s site. Confirm live details on the vendor docs linked in this brief before you quote a client or ship production traffic. Attribute DeepSeek Flash and Pro figures to the 17 September 2026 pricing page when you reuse
-
How cheap is cached input?
Flash cached input: $0.006 peak / $0.003 off-peak per million. Pro: $0.044 / $0.022. That is much cheaper than fresh input. Numbers are from the 17 September 2026 pricing pageu2014recheck live. Confirm live details on the vendor docs linked in this brief before you quote a client or ship production traffic. Attribute DeepSeek Flash and Pro figures to the 17 September 2026 pricing page when you reuse them,
-
Do these prices include Cursor?
No. Cursor Start India and Pro have their own rules. DeepSeek API pricing is what DeepSeek bills when you use their keys. See Cursor docs for editor pools and platform.deepseek.com for API. Confirm live details on the vendor docs linked in this brief before you quote a client or ship production traffic. Attribute DeepSeek Flash and Pro figures to the 17 September 2026 pricing page when you reuse
Sources
Bdeb Technology builds websites, WordPress systems, and tools on top of models like these. A written quote comes back within 24 hours.