Three months ago, DeepSeek publicly promised to turn its promotional pricing into a permanent official rate. Three months later, that rate has gone up fourfold.
DeepSeek announced that, alongside the rollout of its new V4 Pro model, API pricing will change starting August 16, introducing peak and off-peak tiers for the first time — the company says this is meant to "allocate resources more rationally." Take V4 Pro as an example: peak-hour pricing will run $3.96 per million output tokens, more than four times the current $0.87 rate. Off-peak pricing will be half of peak, at $1.98. The lighter-weight V4 Flash model is getting the same treatment, jumping from $0.28 to $1.32 at peak, with off-peak set at $0.66.
What makes this hike feel especially jarring is that the current low pricing was originally billed as a "limited-time promotion," set to expire on May 31. Back then, DeepSeek told the public it would convert the discounted rate directly into its standard pricing — effectively baking "cheap" into a long-term commitment. That promise has now gone unfulfilled, replaced instead by a new price sheet with peak rates four times higher.
Even with the fourfold increase, DeepSeek's V4 lineup still isn't expensive compared to rivals. Chinese company Moonshot's Kimi K3 charges $15 per million output tokens, while OpenAI's flagship GPT-5.6 Sol runs as high as $30. The only option cheaper than DeepSeek V4 Flash's peak rate is OpenAI's budget-focused GPT-5.6 Luna, priced at just $1.2 per million output tokens.
In other words, DeepSeek's strategy of winning market share through low prices is still intact — it's just that the definition of "low" is quietly climbing. The new pricing takes effect August 16, though the specific time windows for peak and off-peak hours weren't included in the materials provided.
