DeepSeek has warned developers that it plans to raise prices across its API services in the near future, saying the increases are expected to be “significant.”
The company announced the change on August 6 but has not disclosed the exact increases or when they will take effect. It advised customers to plan their API usage accordingly.
The move comes less than a month after DeepSeek introduced peak and off-peak pricing. API calls between 9 AM and 12 PM and 2 PM and 6 PM Beijing time cost twice as much during those periods.
DeepSeek’s Current API Prices
DeepSeek currently charges different rates depending on the model and whether input tokens are already cached.
| Model | Input Price (Cache Hit) | Input Price (Cache Miss) | Output Price |
|---|---|---|---|
| DeepSeek-V4-Pro | ~$0.0038 / 1M tokens | ~$0.45 / 1M tokens | ~$0.90 / 1M tokens |
| DeepSeek-V4-Flash | ~$0.003 / 1M tokens | ~$0.15 / 1M tokens | ~$0.30 / 1M tokens |
These rates remain listed in DeepSeek’s official API documentation ahead of the upcoming increase.
DeepSeek Is Still Much Cheaper Than Rivals
Despite the planned hike, DeepSeek currently maintains a substantial pricing advantage over competing AI models.
| Model | Weighted Average Cost for Standard Tasks | Intelligence Index |
|---|---|---|
| DeepSeek-V4-Flash | ~$0.03 | 50 |
| Kimi K3 | ~$0.86 | 57 |
| OpenAI GPT-5.6 Sol | ~$1.86 | More than 59 |
| Anthropic Claude Fable 5 | ~$3.15 | More than 59 |
The comparison also shows that the lower cost comes with weaker overall benchmark performance. DeepSeek-V4-Flash scores 50 on Artificial Analysis’ Intelligence Index, compared with 57 for Kimi K3, while the listed OpenAI and Anthropic models score more than nine points higher.
DeepSeek Moves Away From Ultra-Low Pricing
The increase comes after an intense AI price war in China, where ByteDance, Alibaba, Baidu and other companies have aggressively reduced model prices and offered free services to attract users.
DeepSeek introduced V4-Pro and V4-Flash in April with support for a one-million-token context window. It followed with peak pricing in July before releasing V4-Flash-0731 on July 31 with improved coding and AI agent capabilities.
The latest move suggests DeepSeek is shifting away from its ultra-low-price strategy as AI companies seek more sustainable commercialization models.
However, DeepSeek has not yet revealed how large the increase will be or its effective date. Even after a significant hike, its current price gap with competing models gives it considerable room to remain one of the cheaper major API providers.
The post China’s Cheapest AI Is About to Get a Significant Price Hike appeared first on ProPakistani.
