DeepSeek has begun notifying its clients that the company plans a significant price increase on its API services. The notice, first reported by Bloomberg, gives no specifics on how much rates will rise or when new pricing will take effect, and is framed as advance warning so customers can budget for the change.<\/p>\n
DeepSeek has not disclosed the size of the hike or an effective date. The company appears to be telegraphing the move early rather than announcing hard numbers, suggesting a rollout that could land in the coming months rather than immediately.<\/p>\n
The upcoming change is separate from a mid-July adjustment that raised rates during peak hours in order to ease server load. That earlier move effectively doubled API usage prices in specific time windows, and it remains in place alongside whatever broader pricing update is now in development.<\/p>\n
DeepSeek’s current API pricing on the V4 Flash model is $0.14 per million input tokens and $0.28 per million output tokens. That positioning has made the China-based lab a frequent choice for developers and businesses looking for low-cost inference, particularly after the launch of V4 Flash drew attention for its capability relative to its price.<\/p>\n
Compared with Google, DeepSeek is cheaper on output but pricier on input. Gemini 2.5 Flash-Lite, Google’s budget-tier option, costs $0.10 per million input tokens and $0.40 per million output tokens. Real-world spend depends on workload mix: applications heavy in input tokens, such as long-context analysis or document ingestion, tend to favor DeepSeek, while output-heavy workloads, like content generation, tend to favor Google.<\/p>\n
The timing lines up with a reported infrastructure buildout. DeepSeek has set out plans for a large data center in Inner Mongolia to expand its AI compute capacity, according to a separate Bloomberg report. The company has not officially confirmed the project, but higher API revenue is one plausible way to fund the build, alongside the existing peak-hour surcharge.<\/p>\n
DeepSeek sits in a competitive field alongside OpenAI, Google, and Anthropic, where pricing has generally trended downward over the past year. Any move that narrows the gap between DeepSeek and its Western rivals could push cost-sensitive customers to revisit their provider choices, especially for workloads where the performance gap with top-tier models has shrunk.<\/p>\n
DeepSeek has told clients to expect a significant price increase but has not disclosed specific rates or an effective date.<\/p>\n
No. The July change raised rates during certain peak hours to ease server load, and it remains in effect. The new notice refers to a separate, broader pricing update.<\/p>\n
DeepSeek’s V4 Flash charges $0.14 per million input tokens and $0.28 per million output tokens. Google’s Gemini 2.5 Flash-Lite costs $0.10 per million input tokens and $0.40 per million output tokens, making DeepSeek cheaper on output and more expensive on input. Total cost depends on whether a workload is input- or output-heavy.<\/p>\n