The Chinese company lowered minimum cached input costs to about $0.14 per million tokens, with V4-Pro temporarily discounted 75% further through 5 May to $0.0036.
DeepSeek said the permanent cuts target enterprise clients, developers and agent users, and could spark a price war as rivals including OpenAI, Kimi and Zhipu charge more for flagship models.
Usage rose after V4 launched, with OpenRouter logging 13.6 billion tokens for V4-Pro on 25 April, while integration with Huawei's Ascend ecosystem is expected to improve cost efficiency and adoption.
Is DeepSeek's extreme price cut a sustainable strategy or a cash-burning gamble to capture the market?
Can DeepSeek and Huawei's alliance build an AI ecosystem that rivals Silicon Valley despite U.S. sanctions?
Will ultra-cheap AI models accelerate the rise of autonomous AI agents in our daily lives?
DeepSeek V4-Pro: Delivering Top-Tier AI Performance at 90% Lower Cost Than GPT-5.5
Overview
In April 2026, DeepSeek shocked China's AI market by announcing a 75% discount on its flagship DeepSeek-V4-Pro, enabled by groundbreaking engineering innovations that cut operational costs by 97%. This bold move, supported by a recent partnership with Huawei for seamless hardware integration, disrupted competitors and sparked a rush among enterprises to adopt the affordable, high-performance model. The price cut ignited intense market competition, driving a surge in demand for domestic AI chips amid U.S. export restrictions. While DeepSeek's model slightly trails top global rivals in performance, its dramatically lower cost and open-source licensing make it a game-changer, accelerating AI accessibility, automation, and industry shifts both within China and globally.