DeepSeek V4 Pro API Pricing: $0.66 / $1.98 per 1M Tokens
The numbers: DeepSeek V4 Pro (DeepSeek) bills $0.66 per 1M input tokens, $1.98 per 1M output tokens, and $0.022 per 1M cached input tokens, inside a 1M context window. On our reference 200-step agent workload that works out to about $0.40 per task.
Higher-capability DeepSeek model with thinking mode and 384K maximum output. Every figure on this page was read directly from DeepSeek’s own documentation on September 18, 2026 — not copied from a third-party comparison table. Rates change; the source link and check date below are part of the data.
Verified rates
| Item | Rate | Unit |
|---|---|---|
| Input | $0.66 | per 1M tokens |
| Output | $1.98 | per 1M tokens |
| Cached input (cache read) | $0.022 | per 1M tokens |
| Context window | 1M | tokens |
Billing notes: Off-peak rate shown; peak hours are 01:00-04:00 and 06:00-10:00 UTC on weekdays, when input doubles to $1.32 and output to $3.96 per 1M. DeepSeek had planned to withdraw this model on September 14, 2026 and has since extended it.
Source: DeepSeek official pricing ↗, checked September 18, 2026.
What a real workload costs on DeepSeek V4 Pro
List rates are not bills. These two reference workloads translate the table above into money. The assumptions are visible so you can swap in your own trace — the cost calculator does exactly that.
Scenario A — one 200-step agent task
A coding agent runs 200 steps against a stable 40,000-token prefix (system prompt, tools, repository map). The prefix is written to cache on the first step, re-read on the remaining 199, and each step produces 500 output tokens. For DeepSeek, cache writes bill at the standard input rate.
| Component | Calculation | Cost |
|---|---|---|
| Cache write (once) | 40,000 tokens × $0.66 / 1M | $0.03 |
| Cache reads | 199 × 40,000 × $0.022 / 1M | $0.18 |
| Output | 200 × 500 × $1.98 / 1M | $0.20 |
| Total per task | $0.40 |
Scenario B — one heavy month
A production service pushes 50M fresh input tokens, 150M cached input tokens, and 10M output tokens through DeepSeek V4 Pro in a month:
| Component | Volume | Cost |
|---|---|---|
| Fresh input | 50M × $0.66 | $33.00 |
| Cached input | 150M × $0.022 | $3.30 |
| Output | 10M × $1.98 | $19.80 |
| Monthly total | $56.10 |
DeepSeek rates above are off-peak; weekday peak hours (01:00–04:00 and 06:00–10:00 UTC) double the input and output rates.
Frequently asked questions
How much does DeepSeek V4 Pro cost per million tokens?
DeepSeek V4 Pro costs $0.66 per 1M input tokens and $1.98 per 1M output tokens, with cached input at $0.022 per 1M. Verified September 18, 2026 against DeepSeek's official pricing.
What is the context window of DeepSeek V4 Pro?
DeepSeek V4 Pro supports a 1M token context window, per DeepSeek's documentation checked September 18, 2026.
Does DeepSeek V4 Pro support prompt caching?
Yes. Cached input tokens bill at $0.022 per 1M instead of the full input rate; cache writes bill at the standard input rate.
When was this DeepSeek V4 Pro price last checked?
On September 18, 2026, against DeepSeek's own pricing page: https://api-docs.deepseek.com/quick_start/pricing/