DeepSeek’s latest AI model, V4 Pro, is set to implement a fourfold increase in API pricing, signaling a strategic shift in its pricing model. Beginning August 16, the V4 Pro will cost $3.96 per 1 million output tokens during peak hours – four times the current rate of $0.87 – while off-peak pricing will be reduced to $1.98.
The adjustment introduces a peak/off-peak pricing structure aimed at optimizing resource allocation. Concurrently, the V4 Flash model’s API pricing will rise from $0.28 to $1.32 during peak hours and $0.66 in off-peak periods.
Contrary to earlier plans, the price hike follows the expiration of a promotional period that was initially extended as a permanent discount. Despite these increases, DeepSeek’s V4 models remain more cost-effective than some alternatives. For instance, Moonshot’s Kimi K3 model charges $15 per million tokens, while OpenAI’s GPT-5.6 Sol reaches $30 – with GPT-5.6 Luna offering $1.20 for comparable performance to DeepSeek’s peak-hour pricing.
Leon Neal/Getty Images
