DeepSeek increased its API prices on August 17. The company’s V4 Pro output fee rose from $0.80 to $3.65 per million tokens during peak hours.
Measured drops in volume on some marketplaces
Following the price change, traffic data showed rapid declines on several platforms: on OpenCode, V4 Flash token volume fell 59% on the first day and is now down 79% overall. V4 Pro volume on the same platform has declined by 72%.
Where usage held up — multi-provider marketplaces
At the same time, usage of DeepSeek models remained substantial on OpenRouter, a marketplace where roughly 16 providers compete on price. That pattern suggests developers did not abandon DeepSeek’s models themselves but reacted to pricing: where cheaper alternatives were available, traffic shifted.
Market implication
The episode highlights that AI model buyers are highly price-sensitive; per-token cost became the decisive factor and the market responded within hours to the higher price. DeepSeek appears to offer strong models, but short-term adoption is driven primarily by cost.
Summary of key facts
- Date of price increase: August 17, 2026
- V4 Pro price change: $0.80 → $3.65 per million tokens (peak hours)
- OpenCode: V4 Flash -59% on day one, now -79%; V4 Pro -72%
- OpenRouter: about 16 competing providers; DeepSeek model usage remains large
The immediate takeaway is that in a market with easily substitutable models, pricing determines where usage flows, and a significant price rise can quickly shift demand away on some platforms.
