DeepSeek Extends V4 Pro API Availability Beyond September 14
DeepSeek announced that it will continue offering its DeepSeek V4 Pro API service beyond September 14, 2026, reversing a previous decision to sunset the model. The pricing model for V4 Pro API calls will remain unchanged across both peak and off-peak periods. This extension provides operational stability for developers and enterprises whose pipelines rely on the V4 Pro API, eliminating the need for immediate code changes. It also gives users flexibility to evaluate newer offerings like V4.1 Flash at their own pace. V4 Pro API pricing remains structured around peak and off-peak tiers, with off-peak rates set at 4.5 RMB per million uncached input tokens and 13.5 RMB per million output tokens. Meanwhile, DeepSeek recently released V4.1 Flash, a smaller native multimodal model that surpasses V4 Pro on benchmark evaluation suites at lower API rates.
## BACKGROUND
AI API providers periodically release updated model architectures and plan the deprecation of older generations to optimize infrastructure resources. Additionally, prompt caching is a common API feature where providers charge lower rates for recurring input text prefixes stored in cache memory.