See here: https://api-docs.deepseek.com/quick_start/pricing/

Peak pricing is twice the price, or, off-peak is half the price.

Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC (all other hours are off-peak)

-- V4 FLASH NOW -----------------

Off-peak: $0.007 (cache hit), $0.22 (miss), $0.66 (out)

Peak: $0.014 (cache hit), $0.44 (miss), $1.32 (out)

-- V4 FLASH OLD vs NEW (OFF-PEAK) -----------------

Cache hit: $0.0028 vs $0.007 (+250%)

Cache miss: $0.14 vs $0.22 (+157%)

Output: $0.28 vs $0.66 (+235%)

-- V4 PRO -----------------

Off-peak: $0.022 (cache hit), $0.66 (miss), $1.98 (out)

Peak: $0.044 (cache hit), $1.32 (miss), $3.96 (out)

-----------------

Once the page gets updated and lost in time, to compare, the same page from August 1, 2026: https://web.archive.org/web/20260801170952/https://api-docs.deepseek.com/quick_start/pricing/

I guess it’s not unexpected per se, but the peak hours are solidly Chinese daytime work hours with an extended lunch break, and hitting a bit of Europe’s morning. But no American peak hours, unless you count after-work hobby coding (6-9 pm Pacific?) which tells you indirectly a bit that American adoption is not a major driver of revenue for them. Or that the after-work ‘hobbyist’ is the main audience, which is also plausible.

DeekSeek did a strategic b2c mistake. It is always easier to start with higher prices and give discounts than go opposite way (=raise).

But people get used to higher prices. Or they churn out.

For hobby projects there can be local or free models. There are also some caching techniques.

Well I could have done without the improvement if it meant keeping the prices low, might have to find a better model for cheap inference

Thank you for old vs new comparison