DeepSeek V4-Pro
Last verified September 10, 2026 · DeepSeek pricing ↗$0.660/1M input · $1.98/1M output · 1.0M context · DeepSeek
Count Tokens
—
Tokens
—
Input Cost
—
Output Cost
Estimate Monthly Cost
Monthly Cost Estimator
Quick:
<$0.0001/mo
Pick a preset above or enter custom usage
Alternatives to DeepSeek V4-Pro
Pricing Details
DeepSeekRETIREMENT BY ROUTING, 2026-09-14: DeepSeek's pricing page footnote (2) states that 'From 12:00 Beijing Time on September 14, 2026, and until V4.1 Pro is released in the future, requests to deepseek-v4-pro will all be routed to V4.1 Flash and billed at the V4.1 Flash price.' The scheduled price above ($0.15 / $0.60 off-peak, cache hit $0.003) is therefore NOT a cut to this model; it is the V4.1 Flash card (552B params, 16B active) replacing this 1.6T / 49B one behind the same name. The Pro card itself is unchanged until then. The switch was originally to happen at V4.1 Flash's launch on 2026-09-10 per a user-group notice of 2026-09-09; DeepSeek postponed it four days after a 399-point Hacker News thread. 12:00 Beijing is 04:00 UTC, the minute the first weekday peak window closes. No snapshot id preserves V4 Pro; the postponement notice says continued use after the change is deemed acceptance, with cancellation and refund as the alternative. Whether routed traffic keeps Pro's 500 concurrency or gets Flash's 2,500 is not stated. No V4.1 Pro release date has been published. PEAK IS WEEKDAYS ONLY: the footnote added 'Monday through Friday' between the 2026-08-22 and 2026-08-24 Wayback captures with no changelog entry, so peak covers 35 of 168 hours (20.8%), not 7 of 24; the blended figures below were written before that edit and overstate a flat workload's cost by about 7%. 1.6T total params, 49B active (MoE). GA build DeepSeek-V4-Pro-0813 rolled out 2026-08-13 on the unchanged deepseek-v4-pro model name; self-reported HLE 42.7/60.0 (wo/w tools), Terminal Bench 2.1 87.9, NL2Repo 61.5, Cybergym 83.3, DeepSWE 62.7, Toolathlon-Verified 74.1, Agents' Last Exam 25.7, AutomationBench 31.8, DSBench-FullStack 71.1, DSBench-Hard 67.2, none independently confirmed. Three reasoning modes: non-think, think-high, think-max. Concurrency limit 500 against V4-Flash's 2,500. PRICE CHANGE 2026-08-16: the "significant increase" DeepSeek warned about on 2026-08-06 landed as a peak/off-peak split effective 16:00 UTC on 2026-08-16. Peak is 01:00-04:00 and 06:00-10:00 UTC (7 of 24 hours, the same windows announced in June as 09:00-12:00 / 14:00-18:00 Beijing and then withdrawn); off-peak is exactly half of peak on every line. New card: off-peak $0.66 input / $1.98 output / $0.022 cache hit, peak $1.32 / $3.96 / $0.044. The stored figure is the OFF-PEAK rate, which applies 17 of 24 hours; a workload spread evenly across the clock pays a blended $0.8525 input and $2.5575 output. Multipliers off the old $0.435/$0.87/$0.003625 card are not uniform: input 1.52x off-peak, output 2.28x, cache hits 6.07x, and 3.03x/4.55x/12.14x at peak. The cache discount narrows from 99.17% to 96.67% of the input rate. The output:input ratio moved from 2:1 to exactly 3:1 and the premium over V4-Flash settled at exactly 3.00x. This retires the 75% cut that was made permanent on 2026-05-22 (when the scheduled May 31 step-up to $1.74/$3.48 was cancelled), so "permanent" lasted twelve weeks.
Input / 1M tokens
$0.66
Output / 1M tokens
$1.98
Context Window
1.0M
Max Output
384K
Routed to V4.1 Flash from Sept 14
Scheduled price change
DeepSeek V4-Pro bills at $0.66 input / $1.98 output per 1M tokens through September 13, 2026. From September 14, 2026 the rate becomes $0.15 input / $0.6 output per 1M tokens, a 77% decrease. Every figure on this page uses the rate in effect today.
DeepSeek’s published pricing ↗Price History
| Date | Input /1M | Output /1M | Change |
|---|---|---|---|
| 2026-08-21 | $0.435 → $0.660 | $0.870 → $1.98 | ↑ +52% |
Frequently Asked Questions
Common questions about DeepSeek V4-Pro pricing and usage
Read More About DeepSeek V4-Pro
DeepSeek cut Flash to $0.15 this morning and will bill its Pro tier at the same rate from Monday. For the 96 hours in between, deepseek-v4-pro costs 4.4x more than the model DeepSeek says has already surpassed it.
September 10, 2026 · 13 min read
American teams now route a third of their tokens to Chinese models to save money. The plot twist in July's pricing is that China's best model costs as much as the US flagships it beats.
July 18, 2026 · 8 min read
Meituan's LongCat-2.0 costs 75 cents a million and tops the open coder benchmarks. You can't download it yet, and nobody outside Meituan has rerun the scores.
July 4, 2026 · 8 min read