DeepSeek V4 Prices Jump Over 4.5× – What the Surge Means

One day after the official release of DeepSeek V4, the API price list was updated at 00:00 on August 17, increasing rather than decreasing. The new pricing includes a 30‑day guarantee that subscriptions will not be dropped; if they are, the price difference is refunded on a daily basis.

The standard output prices jumped significantly: Flash rose from ¥2 to a peak of ¥9, and Pro from ¥6 to ¥27, roughly a 4.5‑fold increase. More striking is the cache‑hit price for the Pro tier, which surged from ¥0.025 to ¥0.30 per request—a 12‑fold rise that initially seemed like a data error.

Peak‑Valley Pricing and Off‑Peak Savings

The adjustment is split into two tiers. Beijing‑time 09:00‑12:00 and 16:00‑18:00 are defined as peak periods, while all other times are non‑peak, priced at half the peak rate. After 18:00, the Pro cache‑hit price drops to ¥0.15, and Flash output costs ¥4.5, exactly half of daytime rates.

Consequently, the author advises scheduling batch generation, offline cleaning, and testing after 18:00 to halve costs. Real‑time, high‑traffic services that must run during peak hours should reserve sufficient quota in advance to absorb the higher rates.

For more details, see the new service page at https://shop.apiya.ai.