DeepSeek Switches to Peak/Off-Peak Pricing — With a Hidden Increase
$0.14/1M input, $0.28/1M output (flat rate, V4 Flash)
$0.22/$0.44 input, $0.66/$1.32 output (off-peak/peak) — 1.57× increase
What changed
DeepSeek moved from flat-rate API pricing to a peak/off-peak model on August 16, 2026. But this wasn't just a scheduling change — the base rates went up. The old flat $0.14/1M input became $0.22 off-peak and $0.44 at peak, a 1.57× increase disguised as a time-of-day discount.
Before / After
| Before (flat) | Off-peak | Peak | |
|---|---|---|---|
| V4 Flash input (cache miss) | $0.14 | $0.22 | $0.44 |
| V4 Flash output | $0.28 | $0.66 | $1.32 |
| V4 Flash input (cache hit) | $0.0028 | ~$0.004 | ~$0.008 |
Peak hours: 01:00–04:00 and 06:00–10:00 UTC. Off-peak hours: everything else.
What this means for you
If you were on the old flat rate: your bills went up immediately, even during off-peak hours. The "discount" framing made it look like a scheduling feature, but the cheapest hour now costs more than the old flat rate.
Batch jobs during off-peak still cost less than real-time peak usage — but more than before.
What happened next
DeepSeek reversed course 25 days later with V4.1 Flash. → Sep 10: V4.1 Flash launches at $0.15/$0.60
Related
- DeepSeek pricing — current rates
- DeepSeek vs ChatGPT