Table of Contents
Major Price Restructuring for AI Models
Chinese AI startup DeepSeek will raise API prices for its V4-Pro and V4-Flash models starting August 17, introducing separate peak and off-peak pricing tiers. The new pricing structure will see rates increase by 50% to 1,100% compared with current prices, depending on the model, token type, and time of use.
Peak and Off-Peak Billing Structure
Under the new system, off-peak rates are set at half of peak rates. Peak hours are defined as 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM Beijing Time (01:00–04:00 and 06:00–10:00 UTC), with all other hours billed at off-peak rates.
DeepSeek V4-Pro New Pricing (¥/million tokens)
| Usage Period | Input (Cache Hit) | Input (Cache Miss) | Output |
|---|---|---|---|
| Off-Peak | ¥0.15 | ¥4.50 | ¥13.50 |
| Peak | ¥0.30 | ¥9.00 | ¥27.00 |
DeepSeek V4-Flash New Pricing (¥/million tokens)
| Usage Period | Input (Cache Hit) | Input (Cache Miss) | Output |
|---|---|---|---|
| Off-Peak | ¥0.05 | ¥1.50 | ¥4.50 |
| Peak | ¥0.10 | ¥3.00 | ¥9.00 |
Source: DeepSeek official announcement
Price Comparison and Rationale
For V4-Pro, peak-hour output tokens rise from ¥6 to ¥27 per million (350% increase), while off-peak output tokens increase to ¥13.50 (225% increase). Cache-hit input tokens see the steepest hike—from ¥0.025 to ¥0.30 at peak, an 11-fold (1,100%) increase.
DeepSeek stated the adjustment aims to “allocate resources more reasonably,” and the tiered structure is designed to shift developer workloads toward less congested periods. The company had previously warned of substantial price increases on August 6.
Context of Price Changes
The current pricing was originally promotional, intended to end on May 31. While DeepSeek had announced in May it would make discounted prices permanent, it ultimately decided to proceed with the hikes. The new pricing arrives alongside the general availability of DeepSeek V4-Pro and beta release of V4-Flash.
The price adjustments will only affect API users; web and mobile app users can still access the models for free, including the new “Expert Mode” for V4-Pro.
