← All providers
DeepSeek pricing
verified 2026-09-02Open-weight frontier challenger known for rock-bottom token prices and strong reasoning performance per dollar.
data residency & retention
from DeepSeek's own policies · verified 2026-09-02Headquarters
🇨🇳 China (China)
Where inference runs
China
The privacy policy states personal data is collected, processed, and stored in the People's Republic of China — all API data is processed and stored in mainland China.
Region pinning
No
No region selection, non-China hosting option, or residency commitment appears in the privacy policy, platform terms, or API docs.
Prompt retention
Retained
Prompts are retained as long as necessary to provide services and for interests including service improvement — no fixed window; context caching on disk is on by default with no documented off switch.
Trains on API data
Yes, by default
One policy covers app and API without carve-out: user input is used to train and improve models; opt-out only by emailing privacy@deepseek.com — no console toggle for API traffic.
Zero data retention
Not offered
No zero-data-retention option exists in any official DeepSeek document.
current models · $ per 1M tokens
| Model | Input /1M | Cached in | Output /1M | Context |
|---|---|---|---|---|
DeepSeek V4 Flash DeepSeek V4 | $0.44 | $0.01 | $1.32 | 1M |
DeepSeek V4 Flash Vision (Experimental) DeepSeek V4 | $0.44 | $0.01 | $1.32 | 1M |
DeepSeek V4 Pro DeepSeek V4 | $1.32 | $0.04 | $3.96 | 1M |
DeepSeek totals · 1M in + 1M out
DeepSeek V4 Flash
$1.76
DeepSeek V4 Flash Vision (Experimental)
$1.76
DeepSeek V4 Pro
$5.28
Cost of 1M input + 1M output tokens. Bar length is square-root scaled so cheap models stay visible next to premium ones.