DeepSeek has announced that it will implement price revisions for its Flash series APIs starting from 12:00 Beijing Time on 2026-09-10. During off-peak hours, the unit prices will be $0.003 for input cache hits, $0.15 for input cache misses, and $0.6 for output. Pricing for peak hours is set at twice the off-peak rate.
According to official documentation, current models have been updated to DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813. Additionally, deepseek-v4-flash-vision-exp is being provided as an experimental model supporting image input. DeepSeek's API is compatible with the API formats of OpenAI and Anthropic, allowing users to utilize it by simply changing settings in existing SDKs and tools. Furthermore, a developer preview of "DeepSeek Harness" for agent development has been launched for developers.
Source:
- DeepSeek launching v4.1 flash cheaper and more capable than v4 pro (Hacker News Frontpage, 2026-09-09)
- DeepSeek Official Documentation