DeepSeek extends 75 per cent discount on V4 Pro model until late May
The search engine giant has prolonged its promotional offer on its flagship reasoning model while slashing input cache costs across its entire suite of services
DeepSeek has announced an extension of its promotional pricing structure for the deepseek-v4-pro model, maintaining a substantial 75 per cent discount through 15:59 UTC on 31 May 2026. This move ensures that developers and enterprises continue to access the advanced capabilities of the V4 Pro architecture at a reduced rate for the remainder of the current fiscal period.
Concurrent with the extension of the specific model discount, the company has implemented a broader reduction in operational costs for its API users. From 26 April 2026, the input cache hit price for all DeepSeek models has been slashed to one-tenth of the original launch price. This adjustment applies universally across the platform, significantly lowering the cost base for applications that leverage cached data to generate responses.
In a structural reorganisation of its service offerings, DeepSeek has confirmed that the legacy models deepseek-chat and deepseek-reasoner are scheduled for deprecation. These older endpoints correspond to the non-thinking and thinking modes of the newer deepseek-v4-flash model, respectively. The transition marks a consolidation of the platform's capabilities, directing users toward the more efficient V4 Flash variants for both standard and reasoning tasks.
The billing mechanism for these services remains consistent with the company's standard operating procedures, where charges are calculated based on the total number of input and output tokens. A token is defined as the smallest unit of text recognised by the model, which may comprise a word, a number, or even a punctuation mark. Fees are deducted directly from the user's topped-up balance or granted balance, with the system prioritising the use of granted balances when both are available.
While the current pricing structure offers significant relief to API consumers, DeepSeek notes that product prices are subject to variation and reserves the right to adjust them at any time. The company advises users to regularly check the official documentation page to ensure they are operating under the most current pricing information and to manage their topped-up balances accordingly.
The updates, sourced from the DeepSeek API documentation, highlight a period of rapid evolution in the artificial intelligence sector. By extending discounts and reducing cache costs, the provider aims to sustain high adoption rates for its reasoning models while streamlining the underlying infrastructure for its developer community.
