If your system prompt barely changes, pay cache prices—not full input prices. On the 17 September 2026 pricing page, Flash cached input is $0.006 peak / $0.003 off-peak per million; Pro is $0.044 / $0.022. Fresh Flash/Pro input is much higher. Always recheck platform.deepseek.com for how cache eligibility works live.
The cache rate card
From 17 September 2026: Flash cached input $0.006 peak / $0.003 off-peak per million versus fresh peak input $0.30. Pro cached $0.044 / $0.022 versus fresh peak $1.32. Output still bills at output rates ($1.20 Flash peak, $3.96 Pro peak, off-peak half). Cache does not make completions free. Recheck platform.deepseek.com for exact cache mechanics and minimum token rules. 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. Recheck platform.deepseek.com for Flash and Pro cards, and confirm peak UTC windows
Prompt architecture that caches
Put long, stable policy and tool instructions first. Keep volatile user content at the end. Avoid rewriting the entire system string every call. In PHP or Python workers, load one template and append the job payload. Measure hit rates if the API exposes cache telemetry. India teams scheduling off-peak batches should still structure for cache—timing and cache are separate levers. 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. Recheck platform.deepseek.com for Flash and Pro cards, and confirm peak
What cache does not fix
Cache will not bypass compliance blocks or turn free chat into an API. It will not add vision to Pro. It will not raise Pro concurrency above the cited 500. Use cache after you pick the right model and vendor. Confirm peak windows (weekdays 01:00–04:00 and 06:00–10:00 UTC described) on the pricing page because they can change. 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. Recheck platform.deepseek.com for Flash and Pro cards, and confirm peak UTC windows on
Related on this site
Frequently asked questions
-
Is cached input automatic?
Follow DeepSeek’s live docs for eligibility. This post gives rates from 17 September 2026u2014not implementation guarantees. Recheck platform.deepseek.com. 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.
-
Does cache apply to output?
The cheap cached figures cited here are for input. Output uses output pricing. Cap max tokens to control completion cost. 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.
-
Flash or Pro cache?
Both have cache rows; Flash cache is cheaper. Pick the model for quality needs, then use cache either way. Confirm live numbers. 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.
-
Can Cursor cache DeepSeek for me?
Cursor billing is separate. Other Models on Pro bill at API rates per Cursor docs. Do not assume editor-side cache equals DeepSeek cached input. 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.
Sources
Bdeb Technology builds websites, WordPress systems, and tools on top of models like these. A written quote comes back within 24 hours.