DeepSeek 网摘DeepSeek API Peak and Off-Peak Pricing Explained (October 2026)
DeepSeek's API bills on a two-track peak / off-peak schedule: off-peak unit prices are half of peak. Based on the official pricing page verified on October 16, 2026, this cheat sheet summarizes the current model prices, the peak-window rules, and migration notes.
Current pricing (per 1M tokens)
As of publication, two model names are offered: deepseek-flash (DeepSeek-V4.1-Flash) and deepseek-v4-pro (DeepSeek-V4-Pro-0813).
| Model | Input·cache hit | Input·cache miss | Output |
|---|---|---|---|
| deepseek-flash (off-peak) | $0.003 | $0.15 | $0.60 |
| deepseek-flash (peak) | $0.006 | $0.30 | $1.20 |
| deepseek-v4-pro (off-peak) | $0.022 | $0.66 | $1.98 |
| deepseek-v4-pro (peak) | $0.044 | $1.32 | $3.96 |
Both models support a 1M-token context with up to 384K output, JSON output and tool calls; vision input currently requires Flash, and Pro does not support it.
How the windows work
In UTC, peak hours are 01:00–04:00 and 06:00–10:00, Monday through Friday (excluding Chinese public holidays); everything else is off-peak, including weekends and Chinese holidays in full.
In Beijing time (UTC+8) that maps to 09:00–12:00 and 14:00–18:00 on workdays — the lunch break sits inside an off-peak window. Scheduling batch jobs, evals and data-cleaning runs at night or on weekends halves the bill.
Legacy model names retired
The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but the underlying models are retired: requests are served by DeepSeek-V4.1-Flash and billed at Flash rates. Nothing breaks if you keep the old names, but unifying on deepseek-flash in your next iteration is cleaner.
Concurrency
Concurrency limits are 2500 for Flash and 500 for Pro. High-throughput workloads fit Flash; step up to Pro for harder reasoning and offset the cost with caching and off-peak scheduling (see the cost-saving guide).
Further reading: September pricing update, V4.1-Flash migration guide, launch coverage.