DeepSeek's reasoning-tier model at $0.66/$1.98 per 1M off-peak — a planned reroute to V4.1-Flash's rates was reversed before it took effect.
On September 10, 2026, DeepSeek released V4.1-Flash, retiring the older V4-Flash and V4-Flash-Vision-Exp models (both aliases now point to it) at a lower price than either predecessor. Alongside that launch, DeepSeek said the deepseek-v4-pro API alias would also start routing to V4.1-Flash at V4.1-Flash's rates from September 14, 2026, as a stopgap pending a future V4.1-Pro model. In response to user demand, DeepSeek reversed that plan before it took effect: "we have decided to continue providing API services for DeepSeek V4 Pro after September 14, 2026, with the billing method remaining unchanged," in DeepSeek's own words.
In practice this means V4-Pro's price and capability today are exactly what they were before September 10: $0.66 input / $1.98 output per 1M tokens off-peak, doubling to $1.32/$3.96 at peak. We'll update this page again if DeepSeek revisits the reroute or ships a distinct V4.1-Pro.
V4-Pro (build V4-Pro-0813) is DeepSeek's heavier reasoning-tier model: a 1M context window, 384K max output, and selectable thinking-effort levels (low/high/max) — distinct from and pricier than the lighter-weight V4.1-Flash.
The non-technical considerations are unchanged: DeepSeek is a Chinese company hosting in China, a real factor for regulated western enterprises, and open weights remain available for self-hosting.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| DeepSeek V4-Pro (off-peak) | $0.66 | $1.98 | 1M |
| DeepSeek V4-Pro (peak) | $1.32 | $3.96 | 1M |
| DeepSeek V4.1-Flash (off-peak) | $0.15 | $0.60 | 1M |
| Gemini 3.6 Flash | $0.75 | $3.75 | 1M |
| GPT-5.5 | $5 | $30 | 1M |
With V4-Pro keeping its own, higher price, DeepSeek's real budget leader is V4.1-Flash at $0.15/$0.60. V4-Pro instead competes on reasoning strength per dollar: pricier than Gemini 3.6 Flash on input, but well under a fifth of GPT-5.5's rates on both input and output.