DeepSeek V4 API fully shifts to peak/off-peak pricing on Aug 17, V4-Flash peak-hour price up to 350%

After the V4 family went to general release, DeepSeek officially launched peak/off-peak pricing at midnight on August 17, 2026. Off-peak prices are half of peak prices; peak windows are 9:00 AM-12:00 PM and 2:00 PM-6:00 PM Beijing time, with the rest of the day off-peak.

During peak hours, V4-Pro is ¥9 input (cache miss), ¥27 output, and ¥0.3 cached input per million tokens. Off-peak, V4-Pro is ¥4.5 input (cache miss), ¥13.5 output, and ¥0.15 cached input per million tokens. Versus pre-change pricing, V4-Pro's cached-input price during peak hours rose by up to 1,100 percent (roughly 12x), cache-miss input rose 200 percent, and output rose 350 percent.

V4-Flash pre-change pricing was ¥0.02 cached input, ¥1 input, ¥2 output per million tokens. After the change, peak-hour V4-Flash is ¥0.1 cached input (up 400 percent), ¥3 cache-miss input (up 200 percent), ¥9 output (up 350 percent); off-peak ¥0.05, ¥1.5, ¥4.5 respectively. V4-Flash GA went live July 31 and is the only DeepSeek V4 model supporting Responses API and Codex integration. The price change applies only to the developer-facing API; the consumer web and app experience is unaffected.