- Sources: primary, pricing, discussion, discussion
- Summary: The pricing page states that DeepSeek API billing moves to peak and off-peak rates at 16:00 UTC on 2026-08-16, with off-peak set at half the peak rate, peak hours of 01:00 to 04:00 and 06:00 to 10:00 UTC, and every other hour off-peak. For
deepseek-v4-pro the page lists current rates of 0.003625 dollars per million tokens on a cache hit, 0.435 on a cache miss, and 0.87 for output, against new rates of 0.022 off-peak and 0.044 peak on a cache hit, 0.66 and 1.32 on a cache miss, and 1.98 and 3.96 for output. The deepseek-v4-pro alias now resolves to DeepSeek-V4-Pro-0813. - Comments: A widely upvoted HN comment states that DeepSeek is raising prices starting today. The page dates the increase to 16:00 UTC on 2026-08-16, so the comment is wrong on the date rather than on the rise.
- Why it matters: The cache-hit rate rises about six times off-peak and about twelve times at peak, which falls hardest on workloads built around a warm prompt cache, and callers of
deepseek-v4-pro also changed model version without changing any code. - Follow-up: Confirm whether the peak and off-peak rates take effect at 16:00 UTC on 2026-08-16 as scheduled.
send feedback on this story