DeepSeek's Aug 16 price card is not a 2× peak — Flash output goes from $0.28 to $1.32

DeepSeek's official pricing page now lists peak/off-peak rates effective 16:00 UTC on August 16, 2026. deepseek-v4-flash output becomes $0.66 off-peak and $1.32 at peak, versus $0.28 today. Peak hours are 01:00–04:00 and 06:00–10:00 UTC. Off-peak is half of peak — and still a raise. The same card names DeepSeek-V4-Pro-0813 as the live Pro version.

DeepSeek published the exact peak / off-peak rates on its Models & Pricing page. They take effect at 16:00 UTC on August 16, 2026. (Source: DeepSeek API Docs — Models & Pricing, retrieved 2026-08-14)

Key facts:

Today’s live rates are still $0.14 / $0.28 per million tokens for deepseek-v4-flash (cache miss / output) and $0.435 / $0.87 for deepseek-v4-pro. Cache hits are $0.0028 and $0.003625.
From 16:00 UTC on August 16, Flash output is $0.66 off-peak and $1.32 at peak. Flash cache-miss input is $0.22 / $0.44. Flash cache-hit input is $0.007 / $0.014.
Pro output is $1.98 off-peak and $3.96 at peak. Pro cache-miss input is $0.66 / $1.32. Pro cache-hit input is $0.022 / $0.044.
Peak hours are 01:00–04:00 and 06:00–10:00 UTC. That is 09:00–12:00 and 14:00–18:00 Beijing time. All other hours are off-peak. Off-peak is half the peak rate.
The same table lists MODEL VERSION DeepSeek-V4-Pro-0813 next to DeepSeek-V4-Flash-0731. Context stays 1 million tokens. Max output stays 384K. Both still support thinking and non-thinking modes.

DeepSeek official pricing page showing current Flash $0.14/$0.28 and Pro $0.435/$0.87, plus the new peak/off-peak table: Flash off-peak $0.007/$0.22/$0.66 and peak $0.014/$0.44/$1.32; Pro off-peak $0.022/$0.66/$1.98 and peak $0.044/$1.32/$3.96. Footnote says new prices take effect 16:00 UTC on August 16, 2026. — The official card, retrieved 2026-08-14. Current rates are still in the top table; the new schedule is the footnote table. (Source: DeepSeek API Docs)

What this means if you’re building with DeepSeek V4

1. This is not the 2× peak surcharge the early notice implied. Our V4 Flash guide recorded the earlier plan as 2× during those Beijing windows. The published card is steeper. Flash output goes from $0.28 to $1.32 at peak — 4.7×. Flash cache-miss input goes from $0.14 to $0.44 — 3.1×. Off-peak is not “keep the old price”: Flash output off-peak is $0.66, already 2.4× today’s rate.

2. The cheap cache is still cheap, just less extreme. Today’s Flash miss-to-hit ratio is 50× ($0.14 / $0.0028). After the change it is about 31× at both peak and off-peak ($0.44 / $0.014 and $0.22 / $0.007). Prompt prefixes that stay stable across turns still dominate the invoice. They just no longer buy a 50× discount.

3. Seven peak hours a day are the expensive ones. UTC 01:00–04:00 and 06:00–10:00. Batch evals, overnight agents, and anything that can wait should sit outside those windows. Interactive coding during Beijing afternoon is peak by default.

4. Flash remains the cheaper DeepSeek, even at peak. Peak Flash output is $1.32. Peak Pro output is $3.96. Peak Flash is still below Gemini 3.7 Flash’s $3.75 intro output. The “Flash is the default, Pro is the upgrade” routing in the V4 Flash guide still holds; only the dollars changed.

5. V4-Pro-0813 is now a line on the official price card, not a launch post. DeepSeek has not published a Pro-0813 announcement page. The pricing table is the official name. Treat 0813 as the current Pro checkpoint id, not as a separately documented model drop.

Sources: DeepSeek API Docs — Models & Pricing (retrieved 2026-08-14) · How to use DeepSeek V4 Flash (earlier 2× peak note) · ARC-AGI-2 results for V4 Flash 0731