DeepSeek V4 Pro API prices spike up to 12x starting Aug 17
DeepSeek's new V4 Pro API pricing hits 12x peak rates, with cache-hit tokens costing RMB 0.30...
DeepSeek has officially launched **DeepSeek-V4-Pro-0813** with a dramatic API pricing overhaul effective August 17, 2026. The company is introducing **peak and off-peak pricing tiers**, where off-peak rates are half of peak. For **V4 Pro**, cache-hit input tokens during peak hours (09:00-12:00 and 14:00-18:00 Beijing time) will cost **RMB 0.30 per million tokens**—a **12x increase** from the previous rate. Even off-peak, cache-hit input tokens rise to **RMB 0.15**, a **6x jump**.
The pricing adjustments vary significantly by token type and model. **Cache-miss input tokens** see more modest increases (1.5x off-peak, 3x peak), while **output tokens** rise by 2.25x off-peak and 4.5x at peak. **V4 Flash** follows the same schedule but at lower absolute prices. DeepSeek also emphasizes that the **0813 backend** is now the official V4 Pro model, replacing the Pro Preview build, with no changes required to existing API integrations.
- V4 Pro API prices spike **12x during peak hours** (RMB 0.30/million cache-hit input tokens) and **6x off-peak** starting Aug 17, 2026
- Pricing varies by token type: cache-hit inputs see the steepest hikes, while output tokens rise 4.5x at peak
- V4 Flash follows the same pricing model but at lower rates, and existing API integrations remain unchanged
Why It Matters
Developers must re-optimize workloads to avoid cost shocks from DeepSeek's pricing overhaul.