DeepSeek’s AI models are about to cost four times more

← Back to the feed

DeepSeek’s AI models are about to cost four times more

Engadget · 3 hours ago

DeepSeek is set to raise the API prices of its new V4 Pro and V4 Flash AI models by roughly four times during peak hours from 16 August. The change matters because DeepSeek built its reputation on being substantially cheaper than Western rivals, although its services will generally remain competitively priced.

V4 Pro will cost $3.96 per million output tokens at peak times, up from $0.87, and $1.98 off-peak; V4 Flash will rise from $0.28 to $1.32 peak and $0.66 off-peak. DeepSeek says the peak and off-peak structure will help allocate resources, and notes its previous discounted rates were initially intended as a temporary promotion; even after the increase, V4 models remain below Moonshot’s Kimi K3 at $15 per million tokens, though OpenAI’s lower-cost GPT-5.6 Luna is $1.20.

  • DeepSeek’s new V4 API prices rise sharply from 16 August
  • Off-peak usage will cost half the peak-hour rate
  • DeepSeek remains cheaper than many premium AI rivals

Both sides, in good faith

The strongest fair case each way — we don't pick a winner.

The case for

Supporters of the rise argue that temporary launch discounts cannot responsibly become a permanent business model, especially for compute-intensive AI services. Peak pricing can direct demand towards quieter periods, reduce congestion and help ensure reliable access for users who genuinely need capacity, while still leaving DeepSeek broadly competitive. They would say sustainable pricing is necessary to fund infrastructure, model development and long-term availability.

The case against

Critics argue that a roughly fourfold peak increase weakens the affordability promise that made DeepSeek attractive, particularly for smaller developers and businesses with workloads tied to normal working hours. They may contend that users who planned around the advertised low costs face a sharp and disruptive change, even if the initial pricing was promotional. In this view, demand management should rely more on added capacity, clearer advance commitments or less abrupt price adjustments.

AI Technology

Read the full article at the source →