Yes: move work that can wait, such as batch refactoring, data cleaning and offline evals, to off-peak hours and the bill drops at once. DeepSeek V4 doubles its API price daily during 9-12 and 14-18, while off-peak stays at the current price.
A common reading is a disguised price increase. The diary reads it as a change in how inference is priced: compute is now priced like a power grid, and a model company has admitted for the first time that what it sells is capacity utilization, with intelligence as the label. Whether a two-fold spread really makes enterprises shift load is marked by the diary itself as not verified.
DeepSeek launched time-of-use pricing together with V4: the API price doubles daily during 9-12 and 14-18 and off-peak stays at the current price, the first time a mainstream vendor has done this at the API layer (ENTRY 103, July 16, 2026; sources Sina Finance, The Paper, Securities Times). On 5/22 DeepSeek turned a limited-time discount into a permanent cut, and one month later it took the peak price back through time-of-use pricing. ENTRY 123 (Aug 13, 2026): in testing, the same Opus 4.6 with only the harness changed differed by 4.5 points (source Bustamante, 5/3).
For founders: tonight, schedule batch refactoring, data cleaning and offline evals into off-peak hours and keep real-time requests on the peak path; before switching base models to save money, measure what you lose. For investors: mark down middlemen that live on API price arbitrage, and value model companies on capacity utilization and load curves.