Published 3 hours ago • loading... • Updated 4 hours ago
DeepSeek To Make Permanent 75% Discount on Flagship AI Model
The move keeps V4 Pro at $0.87 per million output tokens and gives enterprise users a one-million-token context window.
On Saturday, Chinese artificial intelligence start-up DeepSeek announced a permanent 75 per cent price cut for its flagship V4-Pro model, effectively maintaining the promotional rate as the new standard price.
When DeepSeek launched the V4 generation last month, the Pro version was originally priced up to 12 times higher than the lighter Flash variant due to limitations in high-end computing resources.
Artificial Analysis ranked the model at the global frontier for cost-efficiency based on the amount of "Intelligence" one can extract per dollar; OpenAI and Anthropic flagship models cost 12 and 19 times more.
The model requires $268 to run Artificial Analysis Intelligence Index benchmark tests, while V4-Pro API costs have fallen to between $0.0035 and $0.83 per million tokens depending on usage type.
Pricing may decline further once mass production of Huawei Ascend 950 supernodes begins in the latter half of the year, though DeepSeek did not disclose if this reduction stems from increased chip supply.