DeepSeek V4 Pro Makes 75% Price Cut Permanent — The AI Price War Heats Up
In late May, DeepSeek made a move that is reshaping the economics of artificial intelligence: it made its 75% promotional discount on V4 Pro permanent.
What was supposed to be a temporary offer (expiring May 31) has become the new normal.
The price war numbers
DeepSeek V4 Pro's permanent pricing runs from $0.003625 to $0.87 per million tokens. For context:
| Provider | Input (per 1M tokens) | Output (per 1M tokens) | |---|---|---| | DeepSeek V4 Pro | $0.0036 | $0.87 | | OpenAI GPT-5 | $2.50 | $10.00 | | Claude Opus 4.7 | $5.00 | $25.00 | | Gemini 3.5 Flash | $0.15 | $0.60 |
DeepSeek costs less than 1% of Claude Opus's price for input tasks.
How did they do it?
DeepSeek says V4 Pro was specifically engineered to cut the cost of long-context inference:
- Uses roughly 1/4 of the single-token compute of its predecessor
- Has 1/10 of the memory footprint at very long contexts
- Runs on Huawei Ascend 950 chips, reducing exposure to US export controls and Nvidia dependency
- 1 million token context window
What this means for the market
The era of high-margin AI tokens may be ending faster than anyone expected. DeepSeek is running the classic Amazon playbook: lose margin on price, win the customer, build the moat.
The structural question for Western AI labs is: how can they compete on price while financing $100B+ infrastructure buildouts?
For developers, this is great news: the cost of integrating AI into products continues to drop dramatically. World-class models are becoming increasingly accessible.