DeepSeek Logo. Image: DeepSeek
In a move aimed at simplifying its offerings, DeepSeek has unveiled an updated pricing structure for its AI models, focusing on greater clarity regarding costs associated with token usage. This new framework details pricing for the two major models: DeepSeek-V3, formerly known as deepseek-chat, and DeepSeek-R1, which replaces deepseek-reasoner.
DeepSeek’s models operate based on token usage, where a token represents the smallest unit of text—be it a word, number, or punctuation. The company has outlined its rates as follows:
| Model | Context Length | Maximum CoT Tokens | Maximum Output Tokens | Input Price (Cache Hit) | Input Price (Cache Miss) | Output Price |
|---|---|---|---|---|---|---|
| DeepSeek-V3 | 64K | - | 8K | $0.07 | $0.27 | $1.10 |
| DeepSeek-R1 | 64K | 32K | 8K | $0.14 | $0.55 | $2.19 |
The company notes, “If the maximum tokens aren’t specified, the default output length stands at 4K. Users are encouraged to adjust their request parameters to accommodate longer responses.”
A significant addition to the updated model is the reasoning component in DeepSeek-R1, termed Chain of Thought (CoT). This feature allows the model to provide reasoning content before delivering the final answer, enhancing the contextual understanding and response quality. Notably, the costs associated with the total output from the reasoning process and final answer are priced equally.
DeepSeek emphasizes transparency in its billing practices: “The expense is calculated as the number of tokens utilized multiplied by the respective price for each model and context setup.” Token counts from both input and output contribute to the final billing, with charges being deducted from the user’s balance accordingly.
Furthermore, the company reserves the right to adjust pricing according to market demands and encourages users to maintain their accounts proactively. “We recommend topping up based on actual usage and to check this page regularly for any updates to pricing.”
As DeepSeek continues to enhance its AI capabilities and accessibility, these new pricing models are expected to aid users in maximizing the potential of DeepSeek’s advanced technology while managing costs efficiently.
Visit DeepSeek API documentation for more detailed insights or to begin integrating DeepSeek into your applications.