DeepSeek announced steep temporary price cuts for its freshly released DeepSeek-V4-Pro model, offering developers a roughly 75% reduction in fees as it seeks to sharpen competition with Silicon Valley artificial intelligence providers.
The Hangzhou-based company implemented the promotional rates last week for those accessing its flagship model. The offer runs through May 5, 2026, at 15:59 UTC, according to the company.
Under the promotion, costs tied to input cache misses were reduced to $0.435 from $1.74. Charges for input cache hits were lowered to $0.03625 from $0.145. Output pricing for the model was cut to $0.87 from $3.48.
In addition to the V4-Pro discounts, DeepSeek said it trimmed fees for input cache hits across its family of AI platforms to one-tenth of their earlier levels. The company framed that change as a direct cost reduction for frequent users who submit similar or repeated requests, noting the lower cache-hit fees should materially reduce expenses for that cohort.
The pricing move comes as OpenAI, Anthropic and Alphabet's Google rapidly roll out new models, platforms that the company and industry observers view as costly to access. DeepSeek launched the V4-Pro after months of anticipation and is positioning the promotional pricing as a competitive lever against those U.S.-based rivals.
DeepSeek previously shook up the market with its R1 model last year, a release that triggered competition among providers focused on price. This latest discount campaign follows that pattern, putting sharper emphasis on unit economics and developer access costs.
Summary
DeepSeek has cut prices for its DeepSeek-V4-Pro model by approximately 75% for a limited time, reducing both cache miss and cache hit input costs as well as output fees, and has lowered cache-hit fees across its platform family to one-tenth of prior pricing. The promotion runs until May 5, 2026, at 15:59 UTC.
Key points
- DeepSeek applied a 75% discount to the DeepSeek-V4-Pro model, lowering input cache miss fees to $0.435, cache hit fees to $0.03625, and output charges to $0.87.
- The company cut input cache-hit fees across its AI platform lineup to one-tenth of previous prices, aimed at reducing costs for frequent users with similar requests.
- The move intensifies competition with OpenAI, Anthropic and Google, whose platforms are described as potentially expensive to access.
Risks and uncertainties
- Duration risk - The price reductions are temporary and scheduled to expire on May 5, 2026, which could lead to higher costs for developers once the promotion ends.
- Competitive response - Rivals may adjust their pricing or product access, which could alter cost dynamics for developers and affect platform economics.
- Usage concentration - The cuts disproportionately benefit frequent users sending repeated or similar requests, creating uncertainty about the distributional impact on different user segments.